This release is dedicated to the Riemann zeta function and accompanies my “Meridian of Primes” series on the Riemann Hypothesis.

Back in 2004, while I was still in high school, I attended a guest lecture on RSA by a visiting mathematics professor from Palacký University in Olomouc. I don’t remember his name, but the lecture was so good that it pulled me deeper into prime numbers, cryptography, and the strange question of how mathematics can describe the distribution of primes.

What fascinated me at that point was that even after so many years, we still do not fully understand the pattern behind them (Riemann postulated the hypothesis in 1859). I filled a few journal pages with notes on the topic back then, and apparently never quite let it go.

At the same time, I was learning about complex numbers in school. They felt bizarre at first. Strange, abstract, almost artificial. And then I discovered that these weird imaginary constructions are deeply connected to the very real, very classical world of prime numbers.

So I decided to connect the topic of primes and complex numbers and start a new series on the Riemann Hypothesis. Not because I plan to solve it, but because I want to explore what it is, why it matters, and how the thinking around it works.

And because Synaptory Fractal Traveler had already grown beyond fractals, this update finally pushed me to rename it. It’s now called Synaptory Chaos Explorer, but the name changes... chaotically :)

Live App

The Riemann Zeta Mode

With version 2.2, Synaptory Chaos Explorer now includes a new mode for exploring the Riemann zeta function (ζ). You can now explore:

• A critical line overlay at Re(s) = 1/2

• The zeta path — the famous visual trajectory of ζ(1/2 + it)

• An analytic continuation for exploring zeta on the full complex plane

And most importantly, the release includes a guided audio-visual tour through 29 curated mathematical points of interest, including:

• Trivial zeros at the negative even integers

• Non-trivial zeros on the critical line, including Gateway Zero, Hardy’s Milestone, or Lehmer’s Phenomenon

• Special values such as the pole, the Basel problem, Apéry’s constant, or Ramanujan summation1.

• Historical landmarks, including Titchmarsh’s last zero or Turing’s last zero

Yes, 29 is a prime number.

Live App

What Else Is New in 2.2

Although this release belongs primarily to the Riemann zeta function, it also pushes the explorer further beyond its original fractal boundaries.

I have added a beta implementation of the Rössler attractor as an early preview of one of my future essays on chaos. It is still unoptimized and experimental, but already stable enough to explore if your machine can keep up. I have also upgraded the Mandelbrot engine with a new series approximation shader, which improves deep-zoom rendering by making extreme magnifications faster, cleaner, and numerically more stable.

Beyond that, version 2.2 adds a new Chaos Gallery for curated captures, smoother tours, richer overlays, expanded keyboard controls, better mobile interaction, internal rendering and architecture improvements, and a substantial set of bug fixes.

So while the zeta function is the headline, this release also marks a broader transition: the project is becoming less a single-purpose fractal viewer and more a growing laboratory for visual mathematics, chaos, and complex systems.

You can find full release notes here.

Give It a Try

• Live app: fractal.brnka.com/#zeta

• Source code & Wiki: github.com/rbrnka/fractal-traveler

Enjoy!

Subscribe now

On the strange atlas of the complex plane, where numbers become geometry and geometry becomes chaos, a unique meridian appears. It marks a boundary between what mathematics has proved and what it has only tested to extraordinary depth.

This essay is a guided walk toward that meridian. Just enough mathematics to see its outline, just enough philosophy to appreciate why it feels so uncanny, and just enough restraint not to pretend we have solved the mystery.

The line is simple, but the claim attached to it remains unproven.

It appears in the Riemann Hypothesis, one of the most famous unsolved problems in mathematics1. The claim is simple to state and extraordinarily hard to prove: all non-trivial zeros of the analytically continued Riemann zeta function lie on that line.

Behind that sentence sits a deeper conceptual shift. To even understand what the claim means, we have to leave behind the naive idea that a formula is only what its original series says it is. We have to talk about convergence, divergence, and one of the strangest moves in mathematics: analytic continuation.

If mathematics has something like an “event horizon,” this problem is one of its best candidates. Yet its proof remains elusive. Proving or disproving it would reshape parts of number theory and would echo far beyond it, including areas tied to prime distribution, computation, and mathematical physics.

You’ll need curiosity more than credentials. Some algebra helps. Comfort with the idea that math can be strange and still be rigorous helps more.

Subscribe now

The Complex Plane

Most people have heard of complex numbers. But what are they really?

We make a mathematically radical move. We say that there is a solution to this simple equation

by introducing a symbol that is not a real number, but whose square is −1. We call it i, define it as

and ask what algebra it forces into existence. The answer is the system of complex numbers, an algebraically closed field, richer than anything that had come before, extending real numbers this way:

Taming Infinity

The most beautiful atlas in the history of mathematics to visualize the extended complex plane2

is the Riemann sphere.

Imagine it as a sphere resting on the origin of the infinite complex plane, at the point zero. We connect every point in the plane to the sphere’s North Pole with a straight line. Where that line pierces the sphere’s surface, the image of the number is born.

Points near zero cluster at the South Pole. But the further you venture into the plane, the higher you climb upon the sphere. All paths to infinity, regardless of their direction, eventually converge at a single point: the North Pole.

This way, Riemann transformed an infinite, ungraspable expanse into a compact, closed object3. Infinity is no longer an abstract abyss; it is a concrete point on a map. And it is within this enclosed world, where algebra meets geometry, that the zeta function begins to reveal its deeper meaning.

Rituals of the Infinite Sum

Before we touch the zeta function, we need to talk about a kind of social contract mathematicians make with infinite series.

An infinite series is a promise that says: “If you keep adding, forever, you approach something.” Sometimes this is true in a way that feels natural. If the terms shrink fast enough, the sum settles down to a number. We say it converges. Sometimes it’s false in the weirdest way possible because the sum simply refuses to be a number, and the series diverges—explodes to infinity or keeps oscillating.

Let me begin with one of the simplest examples:

This is a geometric series S, and it only converges in the narrow interval between -1 and 1 described as:

At x=−1, the partial sums oscillate between 1 and 0. At x=1 they grow without bound. Have a look:

How to find its sum within the convergent interval? Instead of trying to add infinitely many terms one by one, we can manipulate it algebraically. Let’s multiply both sides by x:

and subtract from the original sum:

Everything cancels except the leading 1. Now, to get the sum to one side of the equation and the rest to the other:

Summoning the Right Spirit

We have found a closed-form expression for the sum, but only on the interval where the original series converges. The formula itself, however, defines a function on a much larger domain. Outside the convergence interval, that function still has values, but those values are not the sums of the original series:

How is this possible?

We don’t always treat divergence as failure. Sometimes we treat it as a coordinate change: a hint that we’re summoning the right spirit with the wrong ritual.

Let’s dive in and express the new formula as a function of a real variable x:

We intuitively feel that the function g(x) is somehow extending the S because while equivalent, it works on a broader domain. Let’s plot them both:

We can see that g(x) agrees with the series on the interval where the series converges, but remains defined on a wider domain4.

This is the basic idea behind analytic continuation: the series defines a function on one region, and that function can sometimes be extended beyond it.

Over the real numbers, many different functions can agree on the same interval and then diverge elsewhere. In the complex plane, analytic functions are much more rigid: if an analytic continuation exists on a connected domain, it is unique.

Leave a comment

Beyond Convergence

Let’s play with the g(x). For x=1/2, we get the same result as S. As expected, both give the same value inside the interval of convergence:

Let’s try some of the values outside of the S domain.

As we already know from the graph above, S diverges at +1 while g(x) has a pole there5. How about -1?

For x=−1, the series becomes 1−1+1−1+… Its partial sums oscillate between 1 and 0, so the series does not converge in the usual sense. The function g(−1), on the other hand, gives 1/2, which can be viewed as a generalized value associated with that oscillation6.

Here’s where things get uncomfortable. Let’s try a number 2.

At x=2, the S diverges. The continuation doesn’t. They’re built from the same definition, yet one refuses to answer and the other hands you a number.

This is where the distinction stops feeling like a technicality and starts feeling philosophically strange. The continuation isn’t lying. It’s speaking a different dialect of the same truth. Whether those dialects are equally “real” is a question mathematicians and philosophers still argue about over coffee.

Let’s plot them both:

We can see a very different behavior outside the domain of S. Where S does not converge to a sum, its analytic continuation yields a result (except at +1). The distinction between a function and the series that defines it locally is one of the conceptual keys to the Riemann hypothesis.

But what is the semantics of the number we’re getting? It’s surely not a series sum!

What we get is the value of a different object: a function that agrees with the series where the series converges, and continues beyond that region.

Same voice, wider range.

This is the threshold Euler stepped across in the 18th century, when he faced a problem that had resisted solution for nearly a century: the exact sum of the reciprocals of all perfect squares, known as the Basel problem.

If Euler found the heartbeat of these series, Riemann found their DNA.

In the next part, we will see how the Basel problem evolves into the Zeta function, a mathematical prism that refracts the chaotic sequence of primes into our meridian.

Thanks for reading!

Subscribe now

Back in 2001, I got my hands on the first fractal traveler program written in C by Jindrich Novy. There is no mention of this guy on today’s web, but back then, he was my hero. The program was compact, elegant, and worked surprisingly well, allowing very deep zooms on a Celeron CPU running at 500 MHz. You can find it here.

Since then, I attempted to create a similar experience several times but never got far, mostly due to a lack of motivation to dive into the deeper math of perturbation theory required to make it work beyond naive implementations. And sometimes, dreams come true.

So…

Last year, to kick off my chaos theory series, I built the Synaptory Fractal Traveler, a browser-based, accelerated fractal explorer for zooming into the beauty of Mandelbrot/Julia sets. It came with a whole backlog1 of ideas on how to push it further.

This winter, I finally sat down and burned down enough for the next version.

The quality of my notes let me put Claude Code to work this time. I hoped it would help me with some of the math-heavy challenges, but it caused more damage than clarity there. Where it did help a lot was test generation and a big chunk of the front-end work that would otherwise have taken me ages.

The App

The app is still not where I want it long-term. It’s compute-heavy and will happily spin up every fan your machine has, but it can also generate an absurd amount of beauty.

So after ~a year of iteration, I’m releasing a deeper, faster, more colorful, more polished, responsive, browser-ready version of my fractal zoomer for exploration, learning, and pure visual pleasure of mathematical structures.

If you’re into math, chaos, complexity, visual art, or you just like poking at things you don’t fully understand (but that look amazing), take it for a spin. I also dedicated a lot of time to preselecting some of the intricate structures for you and inventing creative names for them.

• Live App: fractal.brnka.com

• Source, Wiki, Release Notes: github.com/rbrnka/fractal-traveler

Coming

On the roadmap for the next stretch of bad weather: Zeta-function experiments and strange attractors. I’ve actually already rendered the Riemann zeta function in this version of the app (including its analytic continuation), but it’s hidden behind a secret hotkey because it’s not ready yet. If you want to see it, DM me.

Don’t forget to check out the original article on chaos paired with the app:

Enjoy!

This release is dedicated to Jindrich Novy, the author of the original Fractal Traveler from the year 2000, who has inspired me for years to build my own!

Subscribe now

Designing the Interactive Video Player

I loved the idea from the very beginning.

Building a standalone interactive video player within the technical limits of the time—something that genuinely combined a movie and a game—was an enormous and exciting challenge I didn’t want to miss.

The video player follows a playlist of interconnected video clips until the story reaches an ending. During certain clips, the player is presented with a time-limited set of choices, displayed together with a countdown timer. Based on the selected option—or the lack of one—the video player crossfades into the next clip. That transition may be immediate or delayed, but it always needs to feel smooth.

The following diagram explains the core idea:

|           |===== Choices Visible ====|--|
|            ^ showChoices             ^ hideChoices 
|                                      [ crossfade ]    
|---------------- Clip 1 --------------|-----|
|                                      |------ Clip 2 ------>
cut                                   cut (choice made)

Each choice also affects internal variables, which in turn influence future options and enable different endings.

On paper, these were fairly clear product requirements. Every software engineer knows that initial sense of clarity and optimism—right up until implementation begins, and real constraints of technology and user experience start pushing back.

Tech Stack

At first, we considered hosting all clips on YouTube and using embedded video links to navigate between them. That approach failed almost immediately. The experience was fragmented, immersion was effectively zero, and any form of variable logic or deeper interaction was impossible.

When evaluating technologies that could run smoothly on the web and still allow future porting to mobile platforms, Adobe Flash was the obvious choice for me at the time. I had already worked with Flash; it integrated well with the web and supported mobile deployment. HTML5 video APIs, on the other hand, were still immature at the time (YouTube officially switched its default player from Flash to HTML5 in January 20151).

Despite many shortcomings, prototyping complex interactions in Flash was fast and predictable.

The rest of the stack included HTML, CSS, JavaScript & JQuery, PHP & Nette, MySQL & Dibi, JSON & XML.

I initially used SVN for version control before switching to Git in 2023.

Movie Quality

It was 2011. Computers were relatively slow, and mobile devices were just waking up to the world of high-quality media. The first affordable consumer 4K displays appeared in late 2012, and 1080p was still considered high-end.

Keeping the movie at its original full HD resolution, 60 FPS, and zero compression would have resulted in over 20 GB of data to download into the browser cache. With the target market average internet connection speed of ~15 Mbps in 20122, even segmented downloads would take forever.

That was unacceptable. Every scene had to finish downloading before the previous one ended, allowing assets to load quietly in the background during playback.

The final video format was somewhere “in the middle”:

• Container: Flash Video (FLV)

• Codec: H.264 (AVC 1), 960×540 @ 25 FPS

• Audio: AAC Stereo, 48 kHz

This reduced the entire movie to under 200 MB. The compression ratio was determined by the need to minimize loading interruptions while maintaining good quality. To support this, the movie was split into scenes of varying sizes that could be downloaded incrementally and in advance.

Building the Engine

You can’t just let AI vibe-code an application like this. Not in 2011. Even today, that would not work out of the box without tailoring it to the video content.

There was a shared core logic, but a large portion of the code had to be custom-tailored to the movie itself—its pacing, timing, and imperfections. At the time, there were no AI tools to enhance or even generate footage. Everything was built manually, through iteration after iteration, trial and error.

I strongly believe in iterative development. Throughout my professional career, I’ve consistently pushed my teams toward this mindset: start small, but make sure every iteration ends with a working system.

I jumped straight into prototyping based on early mockups and discussions.

I built it on a clean, modular mediator pattern with the main Controller class that coordinates all the specialized subsystems. The heart of the interactivity is the onFrame method: a hook triggered on every rendered movie frame. By anchoring all logic to the video framerate, the system treated the film itself as the primary clock, ticking every 40 milliseconds. I experimented with using time as the clock mechanism, but even with a small lag, the entire experience was ruined.

Matouš and Víťa recorded a few experimental shots with a friend, who served as the early model for Dana. Using just five clips, I built the first functional prototype and verified that sequential playback worked as intended.

At that point, we knew the concept was viable.

In the next iteration—still without final footage—I added branching choices. The prototype was crude, but functional. Subsequent iterations introduced smooth transitions, background music, multiple-choice mechanisms, and the use of the script.

This was not a system that could be designed top-down. On paper, everything looked reasonable. In practice, every prototype revealed a new way the illusion could break—an awkward cut, a mistimed choice, a transition that felt mechanical instead of cinematic. Each iteration wasn’t only about adding features, but also about identifying where the technology became visible and then removing that seam again.

In total, the original version of Interande took 46 iterations (commits) to build and launch. The 2023 version has taken 151 iterations so far.

The Script

From a technical perspective, the script wasn’t a screenplay—it was a data structure. A declarative description of timing, conditions, and transitions implemented as a JSON tree with the following structure:

...
v28a: {next: "v32"},
v28b: {
   next: "v29b",
   showChoices: 170,
   hideChoices: 290,
   points: 1, time: 2,
   choices: [
      {jump: "v29b", x: 250, y: 450, text: "HANA", insta: true},
      {jump: "v29c", x: 430, y: 450, text: "JANA", insta: true},
      {jump: "v29a", x: 600, y: 450, text: "DANA", insta: true},
   ]
},
...
v37x: {
   next: "=time>4?v37y:v37n",
   music: {
      file: "ticho.mp3",
      volume: 0.5
   }
},
...

What mattered was not what the characters said, but when something could be said, interrupted, or redirected without tearing the experience apart. The script defined intent; the engine had to survive reality.

Subscribe now

Semantic Preloading

Although the application includes loading screens, careful optimization ensured they rarely appeared in practice.

The content is split into 16 parts, resulting in 15 potential internal loading moments corresponding to different scenes (bar, city, encounters, and so on). In the original version, scenes were loaded serially. The latest version loads scenes in two parallel queues until everything is fully downloaded—typically before the first scene even ends.

Loading as a Narrative Failure

I also implemented a metric to track the frequency of loading screens during gameplay. Below are statistics showing how many loadings players encountered per play between 2012 and 2025. They illustrate how average connection speeds improved over time, eventually becoming fast enough that downloads completed ahead of playback for 97 % of players by the end of 2025.

The initial peak reflects early load-distribution issues during peak demand. The long-term trend toward uninterrupted playback is visible even though neither the content nor the servers changed. The gap between 2021 and spring 2023 corresponds to the period when the movie was offline.

Since 2023, parallel loading has had only a marginal impact—modern connections are simply fast enough.

Infrastructure as Storytelling

Initially, video files were served from a folder on my personal student hosting plan, which typically handled about 20 MB of traffic per day. After launch—and several times afterward—we peaked at over 230 GB per hour. This triggered alerts on the provider’s side and prompted them to contact me to resolve the situation and move us to a paid plan:

Sun, Jun 14, 2015, 10:00 PM

Dear Mr. Brnka,

During today, data traffic on interande.com increased to 160MB/s (1280Mbps). We are reaching out to resolve this situation and discuss pricing options, as this throughput highly exceeds your plan's SLA.

anafra.cz

While we expected high demand, we never imagined peaks of over 1,000 concurrent players, each streaming 200+ MB of content, sustained for several days. I had to implement a load balancer immediately while the rest of the team scrambled to secure additional storage backends.

We managed to get storage on the Czech University of Technology backbone for a while, and since 2015, the primary storage has been a web RAM disk provided by anafra.cz in exchange for a link on the web. Additional locations mainly serve as backups. This saved Interande, as the average daily data throughput matched a typical monthly quota of regular internet providers.

The load-balancing logic itself uses weighted distribution to prioritize faster storage while spreading traffic evenly. The downloader also implements retries: if a fragment isn’t available on one host, it automatically falls back to another.

The Narrative Angle

A central narrative decision was how the system should behave in the absence of player input. Traditional interactive systems block progression until a choice is made. This approach was incompatible with the cinematic continuity and immersion we wanted to achieve.

Instead, every decision point defined a default jump executed when the choice timer expired. From a technical perspective, this ensured deterministic progression. From a narrative perspective, it prevented the story from acknowledging the player’s hesitation.

The diagram of a “no choice” flow appears as follows:

|   |== Choices Visible ==|
|                                [ crossfade ]
|---------------- Clip 1---------------|------ Clip 2 ------>
cut                                   cut (no choice made = default jump)

Even pausing is disabled during choice windows to prevent the story from acknowledging hesitation.

This mechanism maintains the temporal flow and prevents breaking immersion. Missed choices become narrative events, but the player is slightly penalized for indecision, as the “default path” leads to the worst ending.

Time became an active force rather than a passive resource. If you don’t decide what drink you want, the bartender decides for you!

Illusion of Seamless Editing

This was one of the most challenging parts of the implementation. Each clip began and ended with characters, objects, lighting, and camera framing in slightly different positions. At the same time, player choices could interrupt a clip at any moment, meaning transitions had to work from arbitrary cut points—while still appearing smooth.

The challenge was to programmatically cut the movie in real time and crossfade quickly enough to hide discontinuities. The problem becomes most apparent when a choice window appears near the end of a clip, leaving too little time for a clean crossfade once the decision is made. This caused a few compromises, especially in the dancing scene.

We chose a white transition for player-initiated choices and applied cubic easing to avoid linear, mechanical fades. In the latest version, crossfades are color-coded: black for default transitions, white to confirm that a player’s choice was registered.

Cinematic Karma

Players earn points by guiding the main character through dialogues and challenges, testing reflexes, memory, and decision-making—often without explicit hints. These points unlock or block narrative paths and determine endings.

The engine tracks two variables throughout the game and evaluates branching rules using a simple lexical analyzer driven by the following regular expression:

const regex = /^=(\w+)([><])(\d+)\?([a-zA-Z0-9-]+):([a-zA-Z0-9-]+)$/;

So, for a jump definition of “if you have more than 7 points after this node, jump to v21; otherwise jump to v20” You’d go with:

v19: {next: "=points>7?v20-a:v20-b"}

If you have 7 or fewer points, Dana walks away, and the game ends early.

Music and Sound Player

Flash’s audio capabilities were very limited, so I created a custom sound library to mix video audio with background music and adjust stereo balance. Music tracks were preloaded with each scene and played on loop.

The sound engine supported logarithmic, V-shaped crossfades with configurable duration. A linear fade sounds wrong to the human ear; a trigonometric curve sounds more cinematic.

A constant-power crossfade implementation looks like this:

const p = Math.min((now - crossFadeStart) / duration, 1);

outTrack.volume = Math.cos(p * 0.5 * Math.PI);
inTrack.volume = Math.sin(p * 0.5 * Math.PI);

if (p < 1) {
   requestAnimationFrame(update);
} else {
   outTrack.pause();
   outTrack.volume = 0;
}

The Limits of Portability

The original architecture accounted for mobile platforms from the start. Shortly after launch, I borrowed an Apple iPad and an Asus Transformer Android tablet and began experimenting with native ports.

Porting a Flash ActionScript project to Android and iOS turned out to be far more difficult than anticipated—and not merely because of tooling. Each platform imposed constraints that directly conflicted with the experience we were trying to preserve. Even basic assumptions, such as consistent resolution, aspect ratio, and video rendering behavior, broke down across devices.

Android

On Android, the most immediate limitation was the 50 MB application size cap. This made it impossible to bundle video assets directly into the APK. Instead, the app had to rely on downloading a separate OBB file containing the movie data.

In theory, this mirrored the web model. In practice, mobile connectivity at the time was significantly less reliable. Loading interruptions became frequent enough to disrupt the cinematic flow. To compensate, the entire asset bundle would have to be downloaded upfront and stored locally on the device before the first play. That would take a long time and most likely several attempts due to interruptions and retries.

While this approach was technically feasible, new issues emerged. During testing, I began observing intermittent blackouts during video playback—sudden failures that halted progress entirely. These issues were difficult to reproduce and even more challenging to eliminate reliably across devices.

iOS

iOS presented a different set of problems. Distribution constraints were less severe, but the platform struggled with seamless video transitions. Cuts that were effectively invisible on the web became noticeably abrupt on iOS, breaking immersion at precisely the moments where continuity mattered most.

This wasn’t a cosmetic flaw—it undermined one of the core design principles of the entire project. Smooth transitions were foundational. Without them, the experience no longer behaved like a film.

Why the Web Won

At that point, it became clear that these were not issues that could be solved through incremental fixes. They were fundamental mismatches between platform behavior and narrative requirements.

Rather than ship a compromised version of the experience, we decided to step back from native mobile apps and focus on the web, where the system behaved predictably, and the experience could remain intact.

At the time, expectations were high. The mobile version was meant to launch globally with an English voiceover and serve as a paid product that would help fund future work. That never happened.

More than a decade later, I finally restored the forgotten English assets in the 2023 web release—closing a long-postponed loop and offering Interande to a worldwide audience.

The Web Cinema

The web version of Interande runs on the Nette backend, which handles session, navigation, language support, scoring, and telemetry. The server runs three presenters and two data models.

Every completed playthrough is stored in a database, allowing long-term analysis of player behavior and system performance.

From the beginning, the goal was to keep the experience lightweight. The application needed to load quickly, run reliably, and stay out of the way of the story. There was no interest in visual excess or interface novelty—only in preserving immersion.

During the design phase, Matouš and I settled on a restrained visual language: a dark theme, subtle patterns, a limited color palette, and gentle blur effects. The interface was designed to feel present when needed and invisible otherwise.

The web itself was deliberately cinematic. In the latest version, if the mouse cursor remains inside the video player window, the interface fades out entirely, leaving only the title visible. Controls reappear on pause or when the cursor leaves the video player area—a simple interaction model that reinforces focus on the film rather than the application.

Some Challenges Persist

Safari, in particular, remains problematic due to its restrictions on autoplaying audio and video. Requiring explicit user interaction for every playback transition undermines the very continuity the system is designed to protect. These limitations are outside the application’s control, but they highlight how fragile immersive experiences can be when platform policies intervene.

Despite this, the web remains the most natural home for Interande. It offers reach, flexibility, and—crucially—the ability to evolve. Unlike native platforms, the web version can adapt incrementally without forcing the story into a different shape.

I still have one major task remaining in the TODO list: implementing the correct mobile styling.

Leave a comment

Playing It Again and Again

There were no unit tests. There was no automated test suite. There wasn’t even a meaningful way to simulate the experience without actually playing it.

Internal Instrumentation

For my own dev testing, I used a simple debug panel showing mouse coords for proper choice buttons positioning, variable counters, time elapsed, current and next default jump, frame number/total, autochoice switch, language, memory footprint, and a simple progress bar:

This panel made the invisible visible, allowing me to reason about timing and state without breaking immersion during regular play.

Also, a lot of logging like this:

[net] downlink= 10 parallelDownloads= 2
[preload] preloadPack: Uvod media: 7
[preload] preloadPack: Bar media: 15
[telemetry] game started
[xfade] crossfadeTo: - => v0
[ui] togglePlayPauseButton: ALLOWED 
[music] Switching soundtrack to 0.mp3 vol= 0.5
[xfade] smoothFadeOut enabled
[preload] preloadPack: Sedacky media: 9
[playback] pause()
[focus] User switched away
...

helped me to do most of the heavy lifting. Especially, the autoChoice functionality is priceless. When turned on, it emulates random choices instantly or with a slight delay, allowing you to finish the entire movie in a few seconds to see whether everything calculates correctly, etc.

During the 2012 crunch, we didn’t have this functionality, and the entire team played Interande too many times to stay sane.

Based on the logs, I’ve personally played it over 1000 times!

Testing in the Wild

For external testing, I used my own web-based testing system, which I previously developed to test Flash games. It supported access control, versioning, discussions, SWF instrumentation, and invite keys. It was called FleeBee.

In this system, a small circle of trusted “VIP” testers—our friends, schoolmates, and crew members—helped tirelessly. They took different paths, tried to break the flow, intentionally hesitated, clicked too early or too late, refreshed at the wrong moment, and reported anything that felt off. Logs, screenshots, and long messages describing how something felt were far more valuable than stack traces.

Timing, perception, and continuity only revealed themselves through real playback. A technically correct transition could still feel wrong. A choice that appeared on time could still feel late. The illusion either held—or it didn’t.

Every reported issue fell into one of two categories:

• The system broke (a bug, crash, or missing asset)

• The illusion broke (a bad cut, awkward pause, or scene continuity issue)

The second category was more difficult to fix. The main limitation was that we couldn’t redo the scenes; the material was recorded, and we could only edit what we had. That’s where the script supervisor turned out to be critical—one forgotten purse in the present and then missing in two adjacent shots, and there is no way to cover this up programmatically!

Fixes were validated the same way they were discovered: by playing again. Many times. Often late at night. Usually, after thinking that something was already “done.” Each new build closed some seams and exposed others.

In hindsight, this form of testing more closely resembled rehearsals than verification.

Tracing the Story

From the start, Interande collects telemetry—not for monetization or optimization loops, but to understand how the experience behaves in the real world.

For each playthrough, the system records (besides Google Analytics):

• IP address (for coarse geographic and infrastructure insights, also to roughly distinguish unique players)

• Language and game end timestamp

• Score and final values of internal variables

• Path taken through the story

• Number of loading screens encountered

These metrics serve two purposes.

The first is technical validation. Load counts revealed whether preloading strategies worked as intended. Path distributions exposed edge cases that weren’t sufficiently exercised during testing. Sudden spikes in failed or incomplete sessions often pointed directly to infrastructure or hosting issues.

The second is narrative insight. Seeing how players actually move through the story—where they hesitate, which paths are rare, which endings are common—offers a reality check against assumptions made during design. Some branches that felt important during writing turned out to be rarely reached. Others emerged as dominant simply because they aligned better with player behavior under time pressure.

Over time, the data also became an unintended historical record. Changes in average loading counts mirrored improvements in global connectivity. Infrastructure upgrades showed up as smoother playthroughs, with no changes to a single line of narrative code.

In that sense, the metrics didn’t just describe the game. They documented the environment in which the story lived—and how that environment and player behavior slowly evolved around it.

Final Thoughts

Looking back, I realize I wasn’t just building a video player. I was building a system to manage human attention. Every technical decision—weighted load balancing, frame-accurate timing, smooth crossfades, invisible defaults—served the same purpose: to make the technology disappear.

Immersion in interactive film is fragile. A single loading screen, a cut too visible, or a sound transition that feels mechanical is enough to remind the audience that they are inside an application. Much of the work described here exists solely to prevent that moment—to ensure the story didn’t pause, buffer, or wait, but simply continued.

What surprised me most in retrospect is how closely the technical architecture eventually mirrored the narrative intent. The engine powers more than just a video player. Time moves forward. Indecision has consequences. Continuity matters more than control.

In the next and final article of the series, I’ll share the story of the launch, the shutdown after Flash’s death, and the unexpected rebirth—along with a few bonuses.

Stay tuned, and thanks for reading!

Subscribe now

Language, Our Oldest Technology and Possibly Our Biggest Bottleneck

Human language is an incredible invention. A compressive, lossy codec for the mind.
It evolved for survival, storytelling, gossip, and cooperation — not for describing quantum fields, multi-dimensional manifolds, or embedding spaces. Yet here we are, trying to use 50,000 years of biological slang to communicate with systems that never lived in a savanna, never feared a tiger, and never needed to impress a tribe around a fire.

This alone creates tension.

Language is:

• Ambiguous — almost every word has multiple meanings.

• Biocentric — optimized for human perception.

• Slow — you can think faster than you can speak.

• Shaped by culture — and culture is chaotic.

• Compression-heavy — entire models are reduced to single words (“gravity”, “life”).

• Constantly mutating — Gen Z invents new words faster than models can retrain.

The problem is not that language is “bad.” The problem is that we mistake it for reality.

And Now We Give This Noisy Communication Protocol to Machines!

This is where the trouble begins.

LLMs inherit all the structural constraints of human language — because that’s their training substrate. You get all the ambiguity, all the imprecision, all the hidden assumptions baked right into the embeddings.

When you talk to a model, you don’t give it “the problem.” You give it a sentence about the problem, in a lossy format, constrained by centuries of linguistic drift.

This is why prompting is hard. This is why we need entire fields like “prompt engineering.” We’re performing surgery with crayons.

Despite all its flaws, language enables alignment between humans and machines.
It is our Rosetta Stone: a shared symbolic space.

Language lets us:

• Compress complex models

• Transfer abstract concepts

• Generate structured hypotheses

• Build shared mental models

• Map human goals into machine action

The paradox is beautiful: Language is the weakest link — and yet our only link.

Is Our Language Suitable for a Higher Level of World Comprehension?

If humans and AIs ever build a deeper connection, it may require a new kind of language — one that is less ambiguous, more structural, more mathematical. But until then, we must master the language we have. Even our mathematics can prove too human, too simple, too dumb.

Because if we cannot precisely articulate the problem, we should not expect the machine to deliver a precise solution.

Data, LLMs, GenAI, and the Endless Loop of Knowledge

Large language models take the chaotic data and build a structured manifold: a smooth, continuous space of meaning.

They do not “understand.” But they encode meaning in ways that are:

• high-dimensional

• analogical

• relational

• compositional

• emergent

LLMs build a topology of knowledge that is arguably more coherent than human language itself. This is why they appear intelligent. They are performing geometric reasoning over meaning.

But the tradeoff is severe:

• They inherit every weakness of the data.

• They inherit every weakness of the language.

• They hallucinate where the data is thin.

• They produce statistical echoes of our own linguistic biases.

LLMs are powerful — but they are also mirrors.

GenAI is the Surface Layer We Interact With

It’s the part that produces text, code, images, and ideas. What fascinates me here is that GenAI doesn’t just retrieve patterns. It synthesizes, blends, transforms, and abstracts. Sometimes, it even proposes structures we would not have invented ourselves.

Generative models are compressors and decompressors of possible worlds within the boundaries of language.

This is where the golden quadrant becomes real: GenAI allows us to explore the space of plausible models that humans cannot compute manually.

We now have a four-step loop: Data → Model → Generation → New Data.
This is a brand-new epistemic engine to generate knowledge, as I showed at the beginning.

Can This Loop Produce Something Fundamentally New?

The optimistic narrative you often hear goes something like: “Since LLMs recombine language in new ways, they can discover new truths hidden in text.”

But let’s dissect this carefully.

Language is not just a medium but a latent structure of the world.

Over centuries, humans have encoded:

• patterns

• analogies

• causal assumptions

• heuristics

• models

• folk theories

• scientific abstractions

• Language is a compressed archive of human cognition.

When a model explores this space, it can:

• Reveal latent patterns

• Suggest missing connections

• Propose hypotheses hidden in the structure

• Surface correlations we never articulated

• Generalize beyond any individual text source

In this sense, yes — the geometry of language can hide new knowledge.
And models can discover it.

But only under a condition: The knowledge must be implicit in the corpus. LLMs cannot transcend what humanity has already baked into the linguistic manifold. They share some of our blind spots.

The moment we try to go beyond implicit human knowledge, we hit a wall.

We have no corpus written by physics itself.
No corpus written by evolution.
No corpus written by spacetime.
No corpus written by the underlying machinery of reality.

Until we build foundation models trained on non-linguistic data — raw sensory streams, real physical interactions, symbolic scientific models — language will remain a ceiling.

LLMs can rearrange our thoughts. They cannot generate thoughts the universe never encoded.

Not yet.

So, if the LLM hasn’t encountered the question with its answer before or can’t use a calculator, it won’t be able to determine the answer.

It also won’t make you a coffee.

Not yet :)

Models may help generate novel hypotheses, but not new truth. The moment a model outputs something surprising, we must still:

• Validate it

• Embed it into a scientific model

• Test it

• Use it to make predictions

• See if reality agrees

Language alone is insufficient for scientific discovery. But it is an excellent engine for hypothesis generation.

Leave a comment

That alone moves the boundaries of human knowledge.

This is the direction foundation models are heading — models that integrate:

• spatial data

• physical simulation

• biological signals

• multimodal sensory streams

• mathematical structures

• symbolic scientific rules

Only then will AI be able to generate knowledge that does not simply remix human thought. Until then, we live inside the linguistic bubble our species built.

Near Future Prediction

Predicting the future is traditionally a dangerous sport — but also an irresistible one.
So let me sketch a few trajectories that seem almost inevitable given the current data, trends, and incentives. These aren’t prophecies. Think of them as models: imperfect, but useful.

Data Privacy Becomes Fiction

Privacy stops being about “what you shared” and becomes about “what can be deduced about you.” With enough data points — many of which you never knew you emitted — models will reconstruct your preferences, intentions, habits, and vulnerabilities with unsettling precision.

Privacy becomes less of a right and more of an illusion we nostalgically describe to younger generations.

Democratization of Hacking

Hacking used to require expertise. Now, it requires instructions in natural language. Agentic models can already chain actions, write malware, and adapt to defenses.

When such tools become cheap and ubiquitous, the attack surface grows; defense becomes a continuous battle of AI versus AI.

A Flood of Deepfakes

Deepfakes are no longer a novelty — they’re becoming background noise.
Soon, the question won’t be “Is this video real?” but “Does it matter whether it’s real?”
We’ll operate in a hybrid epistemic environment where synthetic and authentic realities blend. Institutions, media, governments, and individuals will struggle to maintain trust.

Reality becomes probabilistic, just like truth in machine learning.

Advanced Phishing and Social Engineering

Phishing evolves into personalized, AI-generated psychological operations. Models will craft perfect imitations of voices, faces, writing styles — tailored to each individual’s psyche. Your future inbox may consist mostly of messages written precisely for you, by systems designed to exploit every cognitive bias known to neuroscience.

Trust itself becomes a vulnerability.

Job Displacement and Skill Atrophy

AI won’t eliminate all jobs — but it will eat the middle: tasks requiring moderate expertise, moderate creativity, moderate responsibility. The low-end gets automated; the high-end scales with AI assistance. Meanwhile, humans risk losing skills we once considered foundational — writing, reasoning, even the ability to sit with an open problem without asking an LLM for immediate closure.

We may become cognitively dependent, outsourcing more than we realize.

Cheap Dopamine and Infinite Content

AI becomes the cheapest source of dopamine humanity has ever built: hyper-personalized entertainment, companions, narratives, worlds, and attention loops.
We already have infinite content. Soon we’ll have infinite content made specifically for each person, optimized for maximal retention.

The line between “personalized experience” and “addiction-by-design” thins rapidly.

The Path Toward AGI

We will see:

• Autogenous AI: systems capable of improving themselves in constrained domains

• Recursive optimization loops

• Foundation models trained on multimodal sensory reality, not just language or pictures

I don’t know if this leads to AGI, but I do know that the pace of capability growth is accelerating. And with scale, architecture, and multimodality converging, the magnitude of potential knowledge approaches something that feels uncomfortably close to infinity from our human vantage point — the singularity.

If intelligence keeps scaling and models begin optimizing themselves, the idea of a singularity becomes less science fiction and more a matter of when, not if. Not as an explosion of consciousness — but as a tipping point where:

• human pace

• biological limits

• linguistic constraints

• and cognitive bandwidth

simply fail to keep up.

Whether this future is utopian or catastrophic depends less on the models and more on our ability to control the change and adapt.

Thanks for reading!

The structure is as follows:

[SLIDE]

The transcribed text.

And optionally, my commentary.

As you’ve probably guessed from the first slide, we’re at the Faculty of Information Technology, so as a small tribute, I framed this slide in the form of a context-free grammar.

The talk will be with a slightly philosophical flavor, which is always a bit dangerous, because the audience can easily tear you apart in the discussion later.

Let’s dive in.

I created this slide with respect to the venue of the lecture, which was the Faculty of Information Technology at Brno University of Technology, my alma mater. Formal languages and automata were among the most inspiring courses of my studies, and displaying the title as a context-free grammar nicely resonates with the lecture’s themes: the importance of unequivocalness and meaning when working with data and machine learning models. This form clearly reduces the ambiguity of the lecture’s title, which is crucial when querying AI systems.

I represented SolarWinds during the conference, and we got a lovely spot for our booth in the hallways of the old monastery, right across from the refreshments, which turned out to be a strategic position to catch hungry bystanders looking for a schnitzel, introduce them the company, and exchange ideas.

With the surge in AI, new perspectives are unfolding in our understanding of the world around us. In the near future, it will become imperative to move beyond age-old paradigms of truth and reality and start rethinking humanity’s problems to harness the power of AI effectively for their solutions. In this session, I aim to navigate you through the fundamentals of what should be our focus and the necessary mental shift required.

The session abstract was provided a couple of months before the session, which always causes you to deliver something a little bit different from what you thought. Sometimes it’s for the better, I believe. This slide wasn’t part of the presentation and was only provided for the conference agenda.

What are we going to discuss today? I don’t really like agenda slides, so I’ll just splash this all at you, burn it into your retinas for a moment, and then we’ll jump right in.

What the audience saw was a brief cognitive shock of the session’s index, displayed for just a couple of seconds but long enough to imprint whatever their brains noticed first.

Truth

I’ll start with a slightly uncomfortable topic: truth. And why talk about truth in the context of artificial intelligence? As humans, when we grow from children to adults, we usually begin with something like “naive realism”: we think that what we experience and see is a direct reflection of objective reality.

To some extent, this model of truth works. But it also has apparent limitations.

At this moment, your brain, still processing the previous slide, instantly suffers a loss of information abundance, preparing you to focus.

If we base truth only on our senses and experience, we easily end up with “truth” that is basically just a consensus. Everyone says it’s true, so it becomes true. We agree with others, and we build truth from what people around us repeat. And of course, everyone brings a piece of themselves into their picture of truth.

In the end, we bake a strange construct: Very noisy, very coloured by bias, and on top of that, binary: either it’s true or it’s not, which suggests that nothing in between can be true, and suddenly the world gets very black-and-white. In science, this kind of truth model doesn’t work. And that’s already built into the scientific method itself; we’re used to looking at truth differently.

You can see I’ve put a tilde before the “objective” word because I treat objective truth as a purely philosophical concept.

So how do we look at truth from a machine’s point of view? We look at data. Today, truth is statistical primarily. We use statistical methods, but from the perspective of computing and AI, truth is agnostic. Humans can only influence it through the data they produce. If the data are biased, then the “truth” we infer from them is implicitly biased as well. If the data are cleaner, the model of truth becomes cleaner.

What we’re left with is some probabilistic model of truth, and that’s what we actually work with in science and informatics.

So it’s not black-and-white. It’s more of a gradient.

The title is displayed at the same position as on the previous slide to maintain topic fluency. This form of extra dimensionality carries through the entire session wherever possible, it’s also establishing the protocol of how the information in the slides will be presented and what’s the “on slide/spoken” ratio.

The “Truth” is purposely presented within green “walls” that squeeze it between these two paradigms. The image in the background represents a half-man, half-machine according to what level of truth is closer to each of them. The tilde in the “~Objective” says that I don’t believe in an entirely objective truth available to us.

An interesting point on the truth I mention is that "Humans can only influence it through the data they produce.” Of course, especially with GenAI models, we can add multiple layers to “improve” and align the output as needed. My initial thought here was based solely on pure inference.

I’m currently working on a whole article about the topic of truth and will share the link once it’s finished.

Subscribe now

Knowledge

And because this talk is about knowledge, not just truth, the question is: how do we get from this probabilistic model of truth back to a model of knowledge? We must reduce everything into models so that we can even discuss it.

If we start at the level of raw data, we humans receive sensory inputs from the world around us. From that data, we begin to form hypotheses. A hypothesis is nice, but if you can’t test it, it’s almost useless, and nobody will really care. So it’s important to test hypotheses with models.

We’re looking for models that fit the data well and, ideally, have predictive power because both our thinking and our brains are fundamentally predictive. If we had to process everything by brute force — like who is sitting where or standing where — it would be impossible; instead, we constantly predict what will be there, then check whether reality matches our prediction.

We build models in science in exactly the same way. From them, we draw conclusions, and these conclusions can be treated as a certain level of probabilistic truth: truth that holds under specific conditions, in a specific context, with some probability. If you imagine this process as a spiral or an endless cycle of learning, we keep feeding new data back into the cycle and continuously deepen what we know about the world. This whole model has two overlapping boundaries. One is our overall potential for knowledge. The second is the complexity of the models we use to test hypotheses, and therefore the limits of what humans can understand

How do we pursue knowledge when faced with this version of truth? We infer it from data by formulating testable hypotheses and creating models with strong predictive capabilities. The conclusions drawn nudge us closer to the truth and become new data, fueling a relentless cycle driven by our desire for discovery with boundaries only determined by the extent of our potential for knowledge — a frontier that can be shifted..

But what are its limits? I elaborate more on this topic in on of my other lectures:

Knowledge Potential over Time

Let me switch to another little model. An image is worth a thousand words.

On the y-axis, we have “potential for knowledge”, and on the x-axis, we have time. You can probably guess there will be a trending curve, showing how our brains, our collective thinking, and technology have evolved over time. Roughly, at least – reality is always messier.

I’ll mark a few milestones.

The first milestone is the knowledge potential of a single individual. There was probably a time in human history when the entire body of “useful knowledge” could fit into one person’s head. I think it was Komenský who used the term “pansophy”, the universal wisdom, meaning a person could grasp almost everything that could be known and used in life.

That moment is definitely gone. Even if it ever existed, it was very brief.

It began with a single human mind, which once could have contained all available knowledge. However, as our understanding of the world advanced faster than our brains could grow, knowledge outpaced our mental capacity. The term “pansophy”—meaning “universal wisdom”—originates from the 17th century, when it was believed one mind could hold all the knowledge of that era.

Then we learned to work in collectives. At first, small groups, tribes, and teams, but as soon as we started collaborating, sharing, and mixing knowledge and experience, our potential shot up. Everyone could know a little bit, and together we could cover a lot more.

Over time, we started sharing knowledge and expanding it collectively. We collaborated to reach wisdom surpassing individual capabilities. We created various groups—packs, guilds, teams, communities, societies—and together, we learned to hunt mammoths, construct bridges, and carry out complex surgeries such as triple heart bypasses.

The third level we’re roughly at now is the network. Under “network,” you can naturally think of the internet, but also of any large group of people who can exchange vast amounts of data, information, and knowledge in real time and work with it and develop it. That’s the “network stage” – today’s internet, our data reality, our information reality.

This stage faces a major issue: too many inputs, too many actors, and too much noise. It’s becoming very difficult to distinguish between real, reliable knowledge and nonsense.

And of course, how else do we fix a technological problem but with more technology?

Ultimately, we’ve built the most complex structures yet: networks. Many entities, spread across great distances, can now access and share knowledge almost instantly. This knowledge is constantly exchanged, developed, and utilized. However, we’re starting to feel overwhelmed by the huge volume of data. Telling valid models from false ones is becoming more challenging.

This highlights our current situation and explains why I see AI as the essential next step in understanding the vast data landscape.

So here is the breaking point we’re approaching – where artificial intelligence comes into play. Especially this year, it’s a huge hype.

I come from the “old AI” days, when AI wasn’t so sexy – over ten years ago. Today, everyone mostly talks about generative models and large language models. I’m a bit old-school, so I won’t go deep into history; otherwise, we’d be here for more than twenty minutes.

What does AI offer us today? It can work with complex, multimodal data. You can feed it video, talk to it in natural language, give it images, give it all kinds of instructions, and it will talk back to you. That’s roughly where we are now.

AI systems trained on all the data we have, capable of processing complex and multimodal queries and interacting with us in our own language, represent the next frontier. This is likely the point at which we surpass the limits of human potential alone.

But the question remains: how much further can it take us?

I started with machine learning in 2008 by winning an AI tic-tac-toe tournament that was part of our “Fundamentals of Artificial Intelligence” course during my studies. It wasn’t true AI, just smart heuristics, but it was so fast that it actually beat other models thanks to a scoring system that awarded points for both results and speed. It had a complexity of O(SIZE²), no recursion, and no minimax, making it very fast. Because of the missing minimax implementation, I did not receive the top rating despite winning. I focused on optimizing for speed rather than calculating the best move for a long time. But it was enough to spark excitement for machine learning beyond the Terminator movies.

Later, I wrote my master’s thesis on an optimization of neural networks and their combination with chaos theory in stock market prediction. That was well before the genAI bubble.

Where we can go from here is probably more of a topic for an offline beer after the talk. You’ve surely heard terms like AGI, artificial general intelligence, or self-improving AI — systems capable of creating their own enhanced versions and offspring. These systems could become so advanced that they might not communicate with us in our language without a “translation layer” from their internal language, which might be more aligned with the structure of reality itself.

Predicting the future is difficult, but terms like AGI, the singularity, and autogenous AIs are often discussed. I believe these systems will need to go beyond human language limits and redefine many current concepts to better model reality than humans ever could.

Let me make a short detour. I’m a big fan of humans. I think people still have a lot to offer and should try to keep up with AI.

You might have heard the term “transhumanism1”: the idea of extending or enhancing our cognitive abilities. That can include anything from genetic engineering and biohacking to projects like Neuralink from Elon Musk. It’s about anything that could push our potential a little closer to that “upper” curve in the future.

The curve of today’s human potential is not straight; it’s a staircase.

In our model, I’ve represented human potential as a constant for simplicity’s sake. Clearly, it’s not; it has been increasing over time. Yet, the pressing question is whether humanity can keep up with AI’s pace. Is it feasible for human and AI capabilities to converge once more? Could we modify or evolve ourselves to extend our own limits? We have avenues to explore. I am certain there is a path forward. But for what cost?

The big question is: what kind of knowledge can we reach with the help of AI?

Let’s look at one more model.

Leave a comment

Models: Predictive Value vs. Complexity

This time, we’ll model “predictive value” or “usefulness” on the vertical axis, and model complexity on the horizontal axis. We’ll split the graph a bit. Here’s the limit of human potential. That nicely divides the space into two areas: on the left, models humans can understand; on the right, models we probably cannot, but where AI could push us a bit further.

On the left will be most of what we already know, the theories we can actually reason about.

On the right, there’s a space that signifies the window of opportunity for AI.

Our minds and networks are limited in the complexity of models they can understand. We interpret the world through models we developed and comprehend, such as meteorology or quantum mechanics. There is a core body of knowledge within human capacity. Yet, there exists another realm of models that can offer useful predictions but are beyond our ability to understand their internal mechanisms.

I’ll add one more boundary: what I call the “bullshit line”. Anything below this line has insufficient predictive value to warrant scientific or technical interest.

The “bullshit line” marks the threshold of a model’s value and essentially separates what can be considered a valid scientific theory or a useful model from what cannot. Anything below this line is, frankly, unreliable. 🙂

Now, what lives where? Down here on the left, we have folk wisdom, proverbs, rules of thumb – simple models that are good enough for everyday life.

If you miss a train because you imagined it leaving later based on some simple model, it’s annoying but not disastrous.

We don’t need AI there.

There are now four quadrants. The bottom left quadrant contains models that are sufficient for everyday life or to get thousands of followers 🙂

Up here, we have models that are valid and useful: the most rigorous scientific theories and the highest-quality knowledge.

In the upper left quadrant, we find models that are validated, thoroughly tested, and used with a high success rate, yet remain within the complexity our most brilliant minds can handle.

On the bottom right, we have the real bullshit: models that can’t be tested, pure hallucinations, ideas people make up on drugs, totally vague, impossible to formulate, impossible to verify, and therefore useless for prediction.

The bottom right quadrant is designated for untestable models and ideas that appear plausible but cannot be formalized or are simply too wild.

So what’s left is a kind of “golden quadrant”: models that surpass human limits but have high predictive value. That’s where I’d place the biggest potential for AI to expand our knowledge.

The upper right quadrant is our ‘golden quadrant.’ It holds the potential for models that are both highly complex—beyond human understanding—and extremely useful. As mentioned earlier, the human potential line isn’t fixed, and there are models within our potential that we haven't yet discovered. AI can play a key role in uncovering these as well.

I want to give a few examples for the first three quadrants. Treading carefully to avoid offense, in the bottom-right quadrant, I’ve included a single example: the simulation hypothesis2, which is reasonably complex but remains untestable so far. Even if proven, it’s unlikely to provide predictive value beyond Murphy’s Laws, like “If we’re living in a simulation, it’s bound to crash at the worst possible time.”

As shown, the lower left contains flat Earth theory and similar nonsense, while the upper areas approach the human limit, with fields like quantum mechanics and advanced physics. You likely noticed I extended the golden quadrant into this region because we haven’t fully explored the upper left area yet. There’s still a wealth of knowledge to discover, even with AI assisting us in analyzing vast datasets and searching for patterns and structures beyond human perception.

The leading theory of particle physics is the Standard Model. It describes all fundamental physical interactions except for gravity.

Yet, the 2023 Muon g-2 experiment confirmed a 5.1σ deviation from its prediction, with a data fit of about 10⁻⁶. This suggests that our current model is significantly inaccurate, leaving ample room for further exploration.

Subscribe now

The Golden Quadrant of AI’s Potential

So, what kinds of things belong in that golden quadrant?

Most of you know common AI use cases, so I’ll just mention a few. For example, detecting “dirty data3”: highly noisy, corrupted, fake, or fabricated data. That’s something we’ll see more and more of on the internet. Data that is too noisy or has no real underlying foundation, errors, or made-up data. Still data – but much less valuable.

Then there’s what we call “dark data4”: data we gather but never examine, as we don’t see immediate value in it. However, there could be hidden insights there – AI can help explore this area.

In many areas of physics, we seek to unify concepts and interactions, reducing the number of independent ideas. AI could assist by discovering patterns and connections we don’t see. It can help find new combinations and meanings. We describe reality using our human categories, but that doesn’t mean AI will necessarily use the same ones. If AI looks at a table full of bottles, it might see something completely different and label it in a way we would never think of. It can detect symmetries and similarities we’re blind to. That’s a major benefit: AI is not anthropocentric. It doesn’t have to view the world through a human lens. Of course, when we force it into a language model, it becomes very human-centric again. But there are other types of foundation models, not just language-based ones.

Another area is causal inference: we often observe only correlation, but in some cases, AI might help us identify causal relationships. Those are the more practical applications.

If we ventured into the “spiritual” realm, we might ask questions like: what is life, what is consciousness, what is infinity, what is reality, how gravity actually works, or whether free will exists. I don’t dare say that AI will soon help us answer these—still science fiction—but there are already thinkers who try to connect hard science with these traditionally philosophical questions.

There are authors who write about this, and we’re starting to see hypotheses that are at least in principle testable. I think the next few years will be extremely exciting in this area. Assuming, of course, there’s no Skynet moment and it all doesn’t end just as it begins.

Let’s hope not.

Near the edge of human potential, there’s a wealth of knowledge within reach that may be obscured by noise or overlooked because we haven’t yet directed our attention there. AI can help us by sifting through noisy data, extracting value from unprocessed data, and tying up loose ends in our thoughts and hypotheses.

It can integrate concepts, discover combinations we’ve never considered, coin new terms, spot patterns we’ve missed, identify symmetries in data to pinpoint similarities in their origins, and discern causality where we’ve only seen correlation.

If we aim to explore further, many themes and concepts lie outside our current scientific understanding, and our minds may not even grasp them.

So where are we now in terms of our potential, and where does AI sit in that picture?

I borrowed this nice graph from Max Roser, because I really didn’t want to redraw it in ASCII. You can see various human activities that we typically perform quite well, and how AI is catching up or even surpassing us in them.

At first glance, that looks a bit scary, but it also means we’re entering that golden quadrant: AI can now understand written text better than most of us. That’s genuinely scary.

I myself read some books two or three times before I really get them.

How close is AI to helping us with the challenges of the golden quadrant? I’ve adapted this chart from Max Roser in 2023, showing how AI systems have improved over time compared to a human baseline. It demonstrates that recently, AI has started to outperform humans in many areas, so the golden quadrant is not just on the horizon but already here.

Challenges and Paradigm Shift

Let me end with a few challenges we face ahead. Most of them are at the data level.

First: quality. We’ll have more and more dirty data, fake data. How do you build a reliable predictive model if you don’t know what in your dataset is fundamentally correct?

Second: Do we even have enough data? We collect petabytes and exabytes, but is that really enough to understand the universe?

Third: Real-time data. Many models are trained for years; they’re huge and complex, and it’s hard to keep them up to date with fresh data in real time. Often, they sit there and do inference on new inputs, like today’s large language models. We’re not yet good at continuously retraining them to keep up with the latest reality.

Then there’s anthropocentric semantics –I mentioned that already.

Next, we have a lot of “soft data”: if you listen to how Gen Z talks today, which is probably most of you, older people often don’t understand you at all. So why should a machine understand you easily? We use a lot of slang and changing expressions, which is hard for models to keep up with.

The challenges revolve around working with data captured in semantics, which allows us to understand it. Our understanding of data is not needed for AI models, but the data challenges still apply:

• Real-time data presents a significant hurdle, as AI models require extensive time and resources to train, preventing direct real-time processing.

• Soft data, language-specific nuances like narratives and memes, which we sometimes comprehend.

• Noise, dirty data, and false data further complicate matters.

• A key question is whether we have enough data to drive significant progress, and if so, whether we have appropriate computational power to process it.

• Emergent data: New information or patterns that arise at certain levels of complexity must be incorporated.

AI models face inherent challenges:

• Trust in Inductive Reasoning: Can we rely on models if we only have inductive evidence of their validity, knowing they could fail?

• The need for foundational models: We must examine the data on which models are trained and understand their structure. Relying solely on large language models may be insufficient.

And finally: language. I skipped a whole chapter on language, but language is crucial for how we frame problems, communicate them, and share knowledge. We’ll have to learn to ask the right questions and correctly decode the answers. We want answers that are testable – and ideally ones that don’t accidentally blow up a nuclear reactor somewhere.

And that’s it from me. I hope I finished exactly on time.

Thank you.

I skipped the part about the language so I could finish exactly at the 20-minute mark, but at least you get the chance to read it at the end.

Discussion

The last slide I presented was meant to engage the audience and encourage further discussion after the session. Another 15 minutes of questions from the audience followed. Sharing it next. I selected a few questions related to the lecture itself:

Audience: One philosophical question — you touched on transhumanism. Some people connect it with collective consciousness. Isn’t that dangerous?

I see it mainly as an interesting area — we’ll likely have neural implants, augmented cognition… but there are substantial security concerns. With quantum computing coming… …our current cryptographic systems might be broken overnight. So isn’t the risk enormous?

Me: I’ll approach this from two perspectives.

First, I completely agree — connecting humans directly to machines poses significant risks. Bad guys are always one step ahead. You could literally have someone walking around whose thoughts are manipulated externally, and there would be no easy way to detect it.

Regarding cryptography, quantum-resistant crypto is already being developed. We’ll just need a new wave of mathematics and protocols. Honestly, I’m not too worried about this part.

People have always experimented with self-enhancement — mentally and physically. Bryan Johnson takes hundreds of supplements daily, trying to slow aging or improve cognition. Others wear devices that continuously track emotions or measure biometrics, like the first-world cyborg Neil Harbisson. So, this isn’t a new trend. But the more fascinating direction is neuroscience — not enhancing the brain, but understanding it. The brain constructs reality and seems to be a creation to observe and understand. If we know the process of understanding, we might understand reality itself more deeply. That’s why psychedelic-assisted therapy is gaining renewed interest.

Studying how chemicals alter perception allows us to reverse-engineer cognitive patterns. There’s a lot of research happening in this area right now. At the National Institute of Mental Health in the Czech Republic, volunteers line up to participate in live studies, receive psychedelics in a controlled environment, and contribute to ongoing research.

Audience: You mentioned combining science with existential questions— life and consciousness. Have you come across work that does this well?

Me: Yes — I recommend Curt Jaimungal’s podcast “Theories of Everything”. He interviews highly educated guests who present their unique, fascinating models of reality, consciousness, physics, and metaphysics—all thoroughly discussed. If I had to choose one episode, it would be the first one with Bernardo Kastrup, a holder of PhDs in computer science and psychology, who explains his worldview in a very clear, ego-detached way.

Curt Jaimungal

Exploring theoretical physics, consciousness, Ai, and God in a rigorous (and playful) manner.

Audience: You mentioned humans have limits. We do. Does AI have limits?

Me: That’s an excellent question. We don’t know. I skipped a definition of AI in the context of this lecture, but I believe AI can extend our potential and cognitive boundaries, though it may not be the final stage of intelligence.

Everything here is speculative — why would a human-made technology be the end of evolution? The universe is billions of years old; there is plenty of time left, likely for other forms of intelligence to emerge.

I don’t think AI is the final chapter. I also don’t think we’ll be around to see the end.

And there’s more. If you’ve read this far and you’re interested in the additional content, you can find it here:

Thanks for reading, and make sure to subscribe to learn more.

Subscribe now

The furthest scales we have explored—with JWST scanning galaxies, TEMs peering into matter, neuroscience mapping the brain, or submersibles descending into the Mariana Trench—remain many orders of magnitude away from the very fabric of reality.

When we look in both directions—out into the cosmic vastness or inward into the quantum depths—we reach barriers that seem to connect the very large and the very small.

I’ve been contemplating this idea for nearly two decades and have finally decided to share some of my notes. Quantum gravity has always been one of those paradoxes that kept me awake at night or occupied my thoughts whenever I was on the train home, gazing out the window and letting my mind drift. Those were times long before everyone had phones to stare into.

The Macro Horizon: Black Holes

Mass “stores” information via its gravitational field, and when too densely allocated into too small a region, we see it collapse into a black hole.1 Black holes define an event horizon: a threshold beyond which nothing, not even light, can escape, and the information is scrambled in ways we don’t yet understand. From the viewpoint of causality, the horizon is not a material surface but an informational one.

The information paradox of black holes is well known: does information disappear entirely, or is it somehow preserved on the horizon itself, encoded in the faint Hawking radiation?2 The holographic principle suggests the latter—that the universe may be written on surfaces rather than volumes.3 Either way, the event horizon is an informational prison: once crossed, details of the past dissolve into an indistinguishable thermal haze, at least until we solve quantum gravity.

My first encounter with Stephen Hawking’s “A Brief History of Time” was in 2002. I was at the observatory in my hometown of Jeseník, a place where I spent many wonderful years, first as a student and later as a teacher. This book, which was part of the observatory’s collection, ignited my understanding of the cosmos, helping me internalize concepts like distance, size, mass, and gravity. I’ve since moved on, but the book never did—I must admit, I never returned it.

How Far Can We Go?

The heaviest known black hole is likely Phoenix A (at the center of the Phoenix galaxy cluster), with an estimated mass of about 100 billion suns.

It’s located 5.8 billion light-years away and captures an estimated 1.5×10⁹⁹ bits of information. With a Schwarzschild radius of approximately 2.95×10¹⁴ meters, it’s a behemoth 42,000 times larger than our Sun’s radius. Its gravitational pull is so immense that we’re lucky to be so far away from it.

Its event horizon is likely the biggest information horizon in the observable universe.

The Micro Horizon: Quantum Limits

If a black hole’s event horizon is the limit of retrieval, its mirror lies at the other extreme: the limit of resolution where information can’t be distinct from randomness. We model these horizons as holograms, surfaces, strings, foam… Abstractions not far from the two-dimensional event horizon of a black hole.

Heisenberg’s uncertainty principle makes precision in one property erase precision in another. At small enough scales, the act of measurement disturbs the system until outcomes blur into noise.

A signal that approaches pure noise carries no structure. In terms of Kolmogorov complexity, the shortest description is the signal itself — incompressible, indistinguishable from chance.

Both these principles point to a similar conclusion.

How Deep Can We Go?

This “microhorizon” takes several forms: the Planck scale, where spacetime itself may break down; Landauer’s principle, which ties information erasure to minimal energy costs; and quantum decoherence, where superpositions leak into the environment until they can no longer be distinguished. All point to a horizon of measurability, beyond which structure collapses into fog.

At distances smaller than the Planck length

where quantum gravity effects are expected, attempts to probe more precisely would concentrate enough energy to create a singularity (micro black hole), hiding the information again.

The Supersymmetry

Both horizons are limits of distinguishability.

• At the macro scale, black holes scramble information until it cannot be recovered.

• At the micro scale, quantum noise drowns states until they cannot be told apart.

In both directions, entropy saturates, and the effect is the same: information becomes inaccessible.

Both may be expressions of the same holographic encoding at different scales. At the micro scale, trying to measure below the threshold creates a horizon. At the macro scale, piling on mass beyond a certain density creates a horizon.

If these horizons are not separate, then perhaps the universe is a single informational manifold. Large and small aren’t opposites, but projections of the same boundary from different perspectives.

Implications

Distinguishability can be formally captured by relative entropy4, which measures how one probability distribution differs from another. At both the black hole and Planck scale “horizons,” this value approaches zero, which we can write as:

This lack of distinguishability appears to follow a surprising scaling symmetry at both ends of the cosmic spectrum:

• At the macro scale (black holes), the Bekenstein-Hawking entropy (S_BH​) scales with the square of the mass, which in turn is proportional to the surface area of the black hole’s event horizon. This remarkable relationship suggests that the information content of a black hole isn’t proportional to its volume, but to the area of its surface, expressed as:

• At the micro scale (Planck regime), the entropy density (s) scales with the inverse square of the Planck length (lₚ). Because the Planck length represents the smallest possible scale at which spacetime can be measured, this formula implies that the information density at this fundamental level is also a function of a surface-like boundary:

Both follow a quadratic law in surface terms — hinting at a unifying geometry of information, suggesting that the same underlying principles may govern the behavior of information at both the largest and smallest scales of the universe.5

Between Horizons

If these horizons are two faces of a single dualistic boundary, then the real question is not what lies beyond them, but what it means to live between them. We are creatures suspended between these information veils, decoding what structure we can before it dissolves into fog or vanishes into the dark. Our task is not to escape these barriers, but to map the territory they enclose—to learn what order exists inside the finite, and to understand how meaning survives between the extremes of noise and oblivion.

This sharpens the debate about determinism. If randomness is truly fundamental, then entropy is unbounded and the universe indeterminate. If randomness is only apparent—the scrambling of information behind horizons—then chance is ignorance, not destiny. The same logic underpins the question of whether we live in a simulation. Is the information we see emergent and self-contained, or encoded by deeper rules we cannot yet touch?

Full entropy except for black holes is not observed to be locally achievable. If black holes do not randomize (entropize) the information totally and some information can leak back, then it’s not achievable at all.6

Life Between the Veils

Life is a prime example of a system that locally and temporarily fights against this natural increase in entropy.

Every act of living is a form of information processing. A thought, a decision, or a moment of contemplation is a bifurcation of possibility. It’s what transforms the potential into a specific action, making life forms agents of change. Each action is the integral of initial and final states, a physical manifestation of a choice.

Perhaps life emerges when entropy is at its lowest point in a local system, as the universe seeks to express its full potential. However, it’s a property that can only flourish under specific conditions, and it does not always have the ideal environment to maintain high complexity.

Thank you for reading! I will be exploring life and its connection to information as part of this series. If you want to be part of it or just learn more about information, entropy, or gravity, subscribe!

Subscribe now

If you’re interested in exploring some visualizations I created to complement this article, please visit my GitHub and try out the demo.

What makes an interactive movie actually work?

It’s not just clever code or decent actors—it’s alchemy. A mixture of cinematic craft, compelling choices, strong performances, and a team willing to throw everything into a shared vision.

That alchemy clicked with Interande, which somehow racked up over 2.7 million plays in the Czech and Slovak corners of the internet—a digital pond of just 12 million users at the time. Not bad for a no-budget passion startup held together by duct tape, dreams, and the refusal to quit. Sanity clearly wasn’t our strong suit.

But we were dreamers!

It all began back in 2011, when my friend Matouš Ježek pitched me the idea of creating a new generation of interactive film and testing its potential on modern technology. I knew it would be a purely R&D effort and would require a significant amount of time. But I saw the great potential.

We have known each other since childhood. We shared our love for fantasy, gaming, and technology. We collected gamebooks, explored the magical realms of Dungeons & Dragons, and obsessed over Magic: The Gathering. We studied and lived in the same cities for quite some time. While Matouš continued to pursue a career in game design, I focused on game development.

He brought the idea. I brought the tech. I trusted his experience in storytelling and game design, and he relied on my technical skills to bring it all to life. Our solid shared history and passion for games were the foundation we needed.

At that point, we began discussing the technical options available, the timeline, and how to approach the project in an iterative manner. I will elaborate more on the technical side in the following article.

Subscribe now

Matouš had already teamed up with Pavel Gotthard, our screenwriter and executive producer—a true wordsmith who penned a script full of cheeky, memorable lines still quoted today (“Dám ti do toho vodku, aby ti to s ním líp utíkalo / I’ll throw in some vodka to boost the charm levels!”). With a background in dramaturgy and interactive storytelling, Pavel brought both structure and flair.

We now needed someone to shoot the thing.

That’s where Víťa Procházka came in—a meticulous filmmaker who understood that in interactive media, every camera move matters. He choreographed shots with an eye for continuity, knowing that consistency across story branches would make or break the experience.

The Team

Having Pavel, Matouš, and Víťa lead the media side was a turning point. From scripting to shooting alpha test footage for my early prototypes, we worked as one unit.

You can’t simply separate film-making, design, and the in-app experience. They all need to come together, as well as their owners. Filmmakers need to understand the technical possibilities (and limitations) of the technology and prepare the material accordingly. At the same time, software engineering must account for the complexity of the story, game elements, and the limitations of film material.

You already know that the magic of interactive movies is in the thin balance between the quality of the film material and the technology trying to cover up all the imperfections to create an experience as smooth as possible.

And that was my task!

I was finishing my master’s degree and preparing for the final exam, while adding the creation of the interactive player for Interande to my list of leisurely pastimes.

The schedule? Six months from sketches to final product. While I had time to iterate on the tech, the film crew had just two days to capture most of the footage.

We had our core team, but we were excited to bring on talented actors and a fantastic crew!

Interande’s premise was straightforward yet captivating: a classic pick-up artist scenario set in a club, where Petr—our protagonist—is trying to impress Dana, a journalist procrastinating on her article due by sunrise. Not the most responsible journalist, perhaps, but a great character to fit a movie that’s easy enough to film, direct, and gamify.

For this purpose, we needed a special variety of actors who grasp the concept and the game dimension of the movie, are passionate enough, and willing to rehearse multple branches of the story, improvise, and deliver under intense pressure, embracing the chaotic energy required to film an interactive adventure within the narrowest of time frames.

All of that with a budget best described as “non-existent.”

⭐ The Star

We considered several actresses for Dana, but after a few test shoots, Barbora Mudrová stood out—a young Czech actress with a bright smile and even brighter presence. Pavel once joked, “She was the first real actress I talked to,” and she turned out to be the perfect choice: grounded, expressive, and game for the improvisation-heavy chaos ahead.

You!

Two days before filming, we lost the original voice for the main character, Petr. Chaos and panic as everything was already scheduled! Pavel made a last-minute call through his university contacts and somehow roped in the elusive Pavel Hýža. You never see his face in the movie, and honestly, finding photos of him even in real life is a challenge.

He wasn’t just the voice. He was the camera. Since we had no headgear, Víťa filmed the POV shots by holding the Canon 5D camera in front of Pavel, occasionally using his body as a stabilizer. And when the camera needed to show Petr’s hands, they were Víťa’s!

Hand gestures (“ruce”) were even treated explicitly in the shooting plan:

Among our standout supporting cast were memorable figures like bartender Martin Tlapák, Petr’s ex-girlfriend Alžběta Vaculčiaková,

and Dalibor Buš—the drunken friend who later snagged a prestigious Thalia Award.

Not bad for a cast held together by caffeine and good intentions!

The Crew

Then we rallied anyone we could—friends, classmates, girlfriends, strangers from the internet—as extras and testers. But we also needed a real crew: lights, sound, script, makeup, production. That’s when Víťa and Pavel pulled in their networks, and somehow, a whole film crew materialized.

One Night, One City, One Movie

The main interior shoot took place in Brno’s iconic Two Faces club, which generously let us use the space for free. I stopped by to see the chaos and ended up dancing in one of the background scenes.

The pace on stage was fast, and watching Víťa in his element was incredible! Once the club scenes were done, the team moved outside for exteriors, while I returned to my cave to code.

With most footage wrapped, the final challenge was a high-stakes night shoot across Brno.

It kicked off a day later at 7:45 PM in the Husa na provázku Theatre, where Dana sat down for makeup. From there, it was a nonstop race against time. The crew shot through the night, chasing story branches through busy streets and improvised sets. Every scene had to match continuity across multiple outcomes, and every delay risked daylight ruining the shot.

…

It’s early morning on May 1st, 2012, at the bus station somewhere outside the city. The shooting plan explicitly mentions that the sun rises at 05:24 AM to get the final shot.

And so the exhausted crew, fueled only by adrenaline, coffee, and the thrill of creation, captures this last shot in the quiet, awakening outskirts of Brno, Czech Republic. They’ve managed to film the rest of Interande in a single relentless night.

Meanwhile, I’m conspicuously absent from the group photo, asleep at home, preparing to transform their filmed chaos into interactive magic.

Only the easter egg was missing. So we sneaked it in a few days later. Can you guess where it has been filmed?

Leave a comment

The answer is in the following video clip!

That’s how Interande was born—and how it was filmed.

If you’re curious how the interactive player worked, what broke, and how the project eventually rose from the dead, subscribe and stay tuned. The following article dives deep into the technical guts of the platform and how we pushed interactive media beyond “just a gimmick.”

Thank you for reading!

Subscribe now

The problem is universal, qualitative, and somewhat recursive. At its heart, this question relates to whether we are special because we’re biological observers of the universe or whether ‘observers’ (i.e., conscious intelligences) could emerge from other substrates.

Therefore, it raises the question of whether AGI1 is possible and whether what people possess is more than just a “wetware2” algorithm. If consciousness can emerge purely algorithmically, AGI might one day be conscious.

Without a clear handle on this, how can we hope to harness AI to solve our grandest problems? Answering this question would have a profound impact across philosophy, mathematics, and religion. Assumptions about the mind-body problem, implications for soul, spiritual traditions, etc.

We now possess language models trained on almost all our data. If a distorted reflection of reality, modeled through a mirror of language, is enough to create an AGI, then language has always been the key to intelligence and consciousness.

Turing Test 2.0

We have tests that offer satisfactory statistical confidence in evaluating algorithms, intelligence, or human psychology, such as the Turing test or Searle’s Chinese Room3. If we strive to find consciousness using AI, these tests need to be updated, or we may need to accept there is another class of intelligent beings.

Can a general test for consciousness be found, or is it an undecidable problem, perhaps even an impossible one? Is it a task for logic, theoretical informatics, or philosophy?

Every model of reality eventually discovers a set of unsolved, hard problems, including at least one version of the “problem of consciousness.“

Leave a comment

What do you think?

Ethically, let’s draw clear lines of respect for human (and animal) well-being. We don’t need to dissect living brains to gain insights into consciousness… Should we permit it for conscious machines, though? Do you believe AI might never cross this threshold, or is it just a matter of definition before we give it human rights?

Subscribe now

Just recently, I dug up my old notebook from the early years of this century, long before the “AI boom.” In its pages, I found ramblings about how artificial intelligence could shape humanity’s future by putting her at work to solve our most burning issues. Let me share a brief excerpt from it:

16. červen, 2003

…V říši, kde příroda vládne svrchovaně, se na tapisérii existence odvíjí závod o pochopení samotné podstaty života. Je to běh na trati, kde se možná nevyplácí přírodu porazit, protože by se mohla bránit. I tak se ale lidstvo žene za účelem sestrojení inteligence tak výkonné, že už její vývoj samotný zodpoví mnoho otázek dosud nezodpovězených a její výstupy zodpoví mnoho zbývajících. S tím přijde, jako už mnohokrát předtím, posun k hlubší úrovni poznání, která přinese nový pohled na realitu a nové třídy problémů a paradigmat, o kterých se nám ani nesnilo. Pokud se však podíváme na realitu zrakem stroje, nemine nás surový pohled do zrcadla naší vlastní existence obklopený úžasem, znechucením i strachem…

This would translate to English as:

June 16th, 2003

In a realm where nature reigns supreme, a race unfolds upon the tapestry of existence. It’s a race on a track where it may not be wise to defeat nature, for it could fight back. Even so, humanity strives to create an intelligence so powerful that its very development will answer many unanswered questions, and its outputs will provide solutions to the many of remaining ones. With this will come, as has happened countless times before, a leap to the deeper level of reality and a revelation of a new class of problems and paradigms we have never even dreamed of. Yet, when we peer into the fabric of reality through the lens of a machine, we find ourselves gazing into a raw reflection of our existence surrounded with amazement, disgust, and fear.

Back then, the idea sounded outlandish, a dream too far from reality. It’s 2025 now, and the world is awash with AI-driven headlines. It’s the perfect moment to revisit my scribbles and see how they hold up in the present day.

In this series, “23 Questions for AI”, I want to examine the idea of using AI to tackle the world’s biggest problems. But first, before any problem can be put on the table, we need to begin with a few fundamental questions:

• How can we formalize and describe the problems, or if you prefer, how should we ask AI models the right questions?

• How should we decode and validate the answers without harming ourselves? How can we determine if they are correct or not?

• What does it mean to be “correct,” “right,” or “true”? More importantly, is “truth” really the criterion we should use to validate the value of AI’s outputs?

• What role does language play in all this? Will we require a new generation of priests to interpret the AI outputs?

• What is the importance of data? Is it the whole, clear, and raw reflection of the world around us? How do we find models with the best fit on the data we have?

Join me on this journey to learn how language models resemble our consciousness more than we realize and why understanding the language vehicle and the concept of truth are crucial components of this equation.

Subscribe now

Twenty-Three

You are probably wondering—why 23 and not 42?

In 1900, mathematician David Hilbert stood before the International Congress of Mathematicians in Paris to present 23 problems that he believed were crucial to the development of mathematics in the coming century. Although some of those problems remain unsolved to this day, the ripple effect of Hilbert’s challenge reshaped mathematics in profound ways and spurred advances across number theory, logic, and beyond.

Every era has its own grand tasks. For mathematics at the dawn of the 20th century, it was Hilbert’s 23 Problems1—a blueprint for generations of thinkers who sought to push the boundaries of the known. Today, AI has reached a turning point that feels reminiscent of Hilbert’s era, and the problems to be solved call for an equally bold blueprint.

These are the ambitious questions that, once tackled, will shift our understanding of intelligence, consciousness, society, and the underlying fabric of reality itself. They are the questions that, in solving, propel us into an uncharted realm of possibilities.

The 23 is not meant to be a definitive or final number. Rather, it serves as a framework:

1. To focus collective effort on the hardest, most impactful challenges.

2. To illuminate unexplored paths where AI might catalyze entirely new paradigms of thought.

3. To celebrate a tradition of curiosity and intellectual ambition that stretches back into our history.

We know that new puzzles inevitably emerge as soon as one breakthrough occurs.

Problem Criteria

Not every interesting puzzle rises to the level of a problem. I propose four main basic criteria so we don’t propose bullshit:

• Universality: The question addresses a fundamental aspect of intelligence, knowledge, or reality rather than a niche application.

• Transformative Impact: A solution (or substantial progress) on this question would dramatically shift how we live, think, and frame other scientific and societal questions.

• Testability: Even if the problem is partly philosophical, there should be a theoretical or empirical path to test or validate its solution, progress, or initial hypothesis.

• Ethics: We should establish boundaries that reflect ethical values and respect for life, particularly for testing hypotheses.

I considered explicitness one of the criteria, but I’ve decided to put it aside for now. Forcing the question into a strictly formal structure could be counterproductive and limit the set of possible solutions. Let’s revisit it later.

Your Voice Matters!

Throughout this series, you’re invited to:

1. Propose new challenges, questions, and problems you believe could meet the criteria. Or challenge the criteria themselves!

2. Debate the feasibility or importance of the problems discussed.

3. Contribute to refining the final 23-problem list or the criteria. Perhaps your insights will highlight an angle no one else has considered.

4. Challenge!

Ultimately, this series aspires to crowdsource a grand vision that ties together the genius and curiosity of thousands of minds, just as Hilbert’s Problems rallied the mathematicians of the 20th century. Each problem we propose paves the way for new discoveries, fresh paradigms, and deeper insight into the nature of intelligence, ourselves, and our reality.

Leave a comment

Perhaps even some of the unsolved Hilbert’s problems will appear in the list we will build together…

When you think, for example, of a living city and inspect its pulse, you can see some regularities, periodicity, or patterns, but not all of them. You miss the context. You have no idea how the city works because it’s a dynamic system of individual trajectories of everybody and everything there. But if you could understand everyone’s path, the chaos would slightly shift toward order. One step closer but still missing information hidden deeper in the minds of every individual to understand the motivations and processes behind their decisions. Then another step. Their brains, the atoms they are made of, go down and down until you ultimately end up at a dead end of physics, with God or some other level of uncertainty. There will always be a little unknown left, but you will see many symmetrical patterns and self-similarities on the way.

This resemblance of chaos starts at the threshold of predictable structure towards unpredictable complexity. It is ultimately the level of order that transcends our simulation capabilities, meaning predicting the future state of such a system is impossible.

The Essence of Deterministic Chaos

I spent several years commuting by train, traveling to and from school for three hours daily. The journey took me through diverse landscapes and lush forests, where I often admired the intricate shapes crafted by nature. As I spent so much time looking out the window, I began to understand that the repetitive complexity of the structures I saw likely stemmed from something fundamentally simple. This reflection led me to uncover the concept of chaos and the fractal essence of nature. Later on, I studied the work of A. Lyapunov, R. Elliot, or E. Lorenz, and applied concepts of chaos theory in my thesis.

Let’s unveil the mathematical theory of chaos1. It describes non-linear systems that are highly sensitive to initial conditions, where even a minor change can lead to drastically different outcomes. It’s the profane butterfly effect.

Thanks to this sensitivity, a system can start behaving in a way that resembles randomness, but it is deterministic in the sense that it is formally defined, does not contain any random parameters, and if you start with the exact same initial conditions, you will get the exactly same behavior every time.

Even deterministic chaos poses a considerable challenge. You might think that since it is deterministic, it should be manageable with numerical methods. One hallmark of chaotic systems is their high dimensionality. That, in combination with a vast number of possible states, means these systems can shift between various stable configurations through what appears to be random fluctuations, significantly complicating simulations.

A good example is one of the early problems of celestial mechanics and the founding pillars of chaos theory. The Poincaré’s three-body problem. Its idea is to calculate and predict trajectories of three bodies interacting through gravitational force (e.g. three-star solar system) using Newton’s law of universal gravitation. Let me demonstrate the trajectories of these three bodies for you2:

All the bodies have the same mass, zero initial velocities, and only marginal differences in initial vector positions. The rest is done by gravity. This chaotic behavior comes out of three second-order differential equations that don’t have a general analytical solution and are entirely computed numerically.

Fabric of Reality

Now, we understand both notions of chaos. One represents overwhelming complexity, limited only by perception, and the other generates complexity out of simplicity. You surely feel they’re intertwined, and perception and computational capabilities are the keys to their common nature.

Chaotic systems are all around us and shape the reality as we know it. At one end of the spectrum lie simple systems with minimal information – for instance, a mathematical pendulum whose motion can be easily described and predicted. But even Earth’s spin will eventually turn the physical pendulum into a chaotic system.

At the other end are systems containing so many variables and interactions that even the most advanced models can only partially describe them or find limited solutions even for their mathematical representations. Climate, population dynamics, economics3, or even rhythms in biological processes may seem unpredictable, but they have underlying chaos principles at their core.

One notably challenging problem due to its complexity is fluid dynamics. You can imagine it as if you were looking out the window during a snowfall. You can see each snowflake floating beautifully on its unique path; some twirl gracefully with the wind, while others pause for a moment, almost as if they’re enjoying the view before being whisked away again. To truly grasp their behavior, we need an intricate model known as the Navier-Stokes equations. While these equations may explain a lot, they remain an unsolved mathematical problem and keep the prize of one million dollars waiting to be collected.4

Subscribe now

Self-Similarity and Expanding Symmetry

Historically, many mathematicians noticed that iterating on simple formulas in a complex plane could quickly generate unprecedented complexity, and when visualized, actual chaos could be seen! But all early attempts to draw this complexity failed. One of the simplest formulas is

Yes, it is the most famous one—the Mandelbrot set prescription. How does it work? The sequence must not escape to infinity for a complex number c to belong to the Mandelbrot set. Let me show you. Let’s set the initial value of c to a proven constant, like

It is a simple real number within the set, and its sequence converges instead of diverging. Now, to render the first few points in this set without a computer, you would go through this manual exercise:

As you can see, you need very precise decimal point arithmetic just to calculate the first four points. During each iteration, you must check whether the complex number c belongs to the set or not by examining the magnitude (absolute value) of zₙ against this condition:

If |zₙ| is greater than 2, the sequence diverges to infinity, meaning the c value is not in the Mandelbrot set M. It belongs to the set if |zₙ| is less than or equal to 2.5 This binary distinction tells us the basic representation of the point when rendered.

And why 2? It would work for any larger number, but 2 determines the minimal guaranteed escape radius and size of the disk surrounding the entire set.

Let’s use asterisks (*) for points in the set and dashes (-) for points outside the set as Benoît Mandelbrot did in the 1980s when he rendered probably the very first Mandelbrot set using a computer6. After hours of patient waiting, the result in the form of a vaguely ass-shaped blob finally appeared:

This grainy, low-resolution render was a doorway to a world of infinite beauty, where math and art collide—a world of fractals.

Let’s Travel Together!

If you’re curious about what the Mandelbrot fractal may look like today, I wrote a web app allowing you to travel the set. It is limited by the threshold of the 32-bit floating-point arithmetic in your GPU’s fragment shader and can’t go very far, but you can still explore some of the intricate structures. It also contains Julia mode to explore Julia fractal.

~ Synaptory Fractal Traveler ~

The pseudocode of the main method is very simple7:

def mandelbrot(c, max_iterations):
    z = 0
    for n in range(max_iterations):
        z = z**2 + c
        if abs(z) > 2:
            # Normalize the escape value for smooth coloring
            return n + 1 - log2(log2(abs(z)))
    return max_iterations

The values are normalized to achieve fancy coloring (faster escape = brighter colors, slower escape = darker colors). The app allows you to randomize the color palette, and I have also preprogrammed a few interesting zoom-ins you can admire. Let me know your thoughts, or share some of your zoom-ins!

Leave a comment

Compare the difference. Medieval mathematicians could only imagine a few points. In his first render, Mandelbrot saw a few hundred points of the set on a print corresponding to the two orders of magnitude. My app can do seven orders of magnitude before it pixelates. The best fractal zoomers can render the set at more than 20,000 orders of magnitude. Computing power allows us to push the limits and delve into previously hidden structures and patterns of chaos.

But Where Is The Chaos, Exactly?

With increasing precision and the number of iterations, we discover sets of points that do not change their behavior very often during the calculation but also sets of points that are very sensitive to changes. The boundaries between them create the fractals. That’s why the Mandelbrot set has a distinctive shape, and the Julia set changes its look rapidly with a small change of the c at certain points. It is always the coastline where fractal shapes appear.

Fractals are, therefore, another example of deterministic chaos. Even a simple equation can generate unpredictable complexity, yet it is always the same. Fractal geometry reveals that even seemingly disordered structures contain recurring patterns. Discovering its existence transformed my previous commuting insights into a genuine passion. Fractal shapes can be found in tree growth, cloud formations, river flows, or Romanesco broccoli, forming the world around us.

World Without Chaos

Although chaos may initially appear negative, you just learned that it is, in fact, an indispensable source of innovation and change. Systems that are overly rigid and orderly can easily become stagnant. Chaos introduces diversity, new possibilities, and unexpected connections. In evolution, chaos drives adaptation and the emergence of new species. Without it, humankind would probably not exist.

Chaos provides a space for new ideas and perspectives. It challenges us to think beyond the confines of established patterns and explore uncharted territories of thought and imagination. Psychedelic hallucinations often present themselves in fractal shapes, reflecting the structure of our brains and our thoughts.

Chaos forces us to confront the nature of our existence. It blurs the boundary between determinism and free will, suggesting that the universe is a dynamic interplay of patterns and unpredictability at the verge of our perception. In chaos, we find a mirror of human life itself: an intricate balance between order and spontaneity.

Thank you for reading! If you want to learn more about how chaos affects the world around us, what role prime numbers play in shaping reality, or more about fractals, subscribe for free to receive new posts and support my work.

Subscribe now

Forest of Endings

A film delivers a linear experience with a single, predetermined ending. As much as there are interactive movies with a single ending (like the Kinoautomat I mentioned in Part 1), I think more endings hugely improve engagement and repeatability. After all, why stick to a single ending when you don’t have to?

Having numerous endings increases not just the movie’s complexity but also its cost1. Interactive movies face physical limitations when using conventional filmmaking methods.

Tree of Choices

Excessive choices lead to an overwhelming number of branches, necessitating the filming of all possible paths and outcomes.

Look at Chess,

where each of your moves explores its own subtree of possibilities and choices, including the responses your opponent can make and your reactions to those responses. Considering the typical number of legal moves available at a given position is 35 (b), and the typical number of plies is 80 (d), the tree of all these combinations contains up to

nodes. That’s over 40 orders of magnitude greater than the number of atoms in the universe. But even games that can attain numerous internal states frequently reduce almost infinite possibilities to a few potential endings. Chess into win or lose. The interactive film also adopts a graph structure of a decision tree that eventually collapses into a few endings.

When you make a movie with real actors in real environments, it can be cumbersome to get everything into the same setup for multiple choices: items, lights, body positions, and people’s faces, especially!

The greatest challenge with multiple choices is achieving a seamless and smooth transition when cutting the scene.

One thing you should avoid is loops. Loops create a sense of déjà vu. Each moment in the movie must be unique!

Interande: The Snake of Choices

Interande is also designed as a decision tree, shaped into a linear, snake-like shape thanks to the flow of time. It includes up to 28 decision points and 5 different endings. I’ve visualized the structure for you:

The red node represents the start, the blue nodes are endings, the green nodes are decision points, and the grey nodes represent regular scenes (cut scenes). There is one unique checkpoint leading to the first ending that decides whether you make it to the main part of the movie. Can you find it?

The gameplay contains internal variables that shift with the player’s choices, allowing for multiple endings through conditional decision nodes. This system unlocks choices that might otherwise be inaccessible and opens paths that would otherwise stay hidden, enhancing the complexity and personalization of the experience. The two variables tracked globally in Interande are the ❤️Chemistry Index and the ⌛Runaway Index. I’ll share what they do later.

Subscribe now

In Interande, you typically make a choice every 20 seconds, choosing from up to 3 options. So, for a 15-minute experience, there are around 35 minutes of video content, represented by 113 nodes in the snake. However, the shortest path through the movie requires only 57 scenes to reach an ending, while the longest path consists of 66 scenes. There are no loops, no checkpoints, and no save/loads.

You can compare it with the decision tree of Black Mirror: Bandersnatch, which, according to Netflix, contains as many as five endings, 31 decisions, and no variables.

The cut scenes are very long, and there is unlimited time to make choices, making the experience quite different.

Run Forest, Run!

Every movie must maintain its pace, inherently imposing time constraints on the player and enhancing interactivity. In the real world, your date won’t pause for two minutes while you contemplate what to say—there are consequences for that. This leads me to the next limitation: time to choose. There are two general ways to handle timed choices to maintain a good experience. 

The first one is manipulating time, using a bullet time or a pause. There are cases where this makes sense for sure. The other tries to maximize the fluency and the immersion; the ingame time must go on, and the scene must be directed to allow the player enough time to decide while keeping the pressure of the surrounding world just continues to live.

This constant pressure actively increases the player’s engagement. Like in life, choices may appear at any moment and quickly disappear. Once gone, one of the choices is usually taken automatically to maintain the pulse—usually a poor one!

Sometimes, however, a hidden pathway leads beyond the presented choices, especially in moments when silence or nothing is the preferred course of action.

Show the Clock?

An intriguing problem is whether and how to visualize the remaining time during the decision. Not telling the player is interesting but likely unsuitable for casual play. In the original Czech-only version, we used a shrinking progress bar at the bottom of the screen,

whereas in the Reloaded version, I redesigned the choices to fade out slowly with the time running out. This doesn’t say how much time the player has, only when it is about to run out:

Which one do you like better?

Leave a comment

Our Niche

Our concept was innovative. I created a scalable, publicly available web application built on a reusable platform, which allowed us to glue movie scenes into the snake and provide the experience of a “point-and-click” adventure. That alone is not enough. Many components must come together perfectly. If you’re curious about which ones,

Subscribe now

Thank you for reading!

Let’s examine the space between the level of control and realism. I don’t imply that a medium adheres strictly to the laws of physics when speaking of realism. It’s a measure of how effectively it immerses the player. By control, I refer to the frequency and extent to which the player can intervene, make choices, or influence the internal reality.

When a medium provides too much control, it resembles a sandbox—a simulation environment without inner pace. If it extends for a prolonged period without interactive elements, it resembles a conventional film.

Similar to video games, interactive films aim for the intersection of sufficient realism and a satisfying level of control, balancing immersion and the film experience in, as I tend to call it, the Immersion Quadrant.

Realism: The Essence of Immersion

The sole point behind realism is to create an experience tricking your brain into generating a response as if you were truly there—whether on a battlefield or in a bar! The goal is to bring your brain as close to reality as possible and keep you there. To squeeze the neurotransmitters out through a well-mixed cocktail of sensory stimuli. Sometimes, the brain response can be so intense that it leads to addiction.1

One of the best-known games of the Immersion Quadrant is Heavy Rain (2010), thanks to its thick narrative, use of actual movie footage to place the story into real events, the player’s quick choices throughout the gameplay, and multiple endings.

As a dedicated gamer, I learned about various approaches to immersion from multiple perspectives. Being a game developer helped me gain even deeper insight into these mechanics, and this combination will always keep my internal passion for interactive experiences burning.

The rabbit hole goes even deeper when it comes to immersion in today’s video games. The braindance feature in Cyberpunk 2077 (2020) allows players to relive someone else’s memories in a simulated recording of their own experience.

Soon, computer graphics will achieve a level of realism that, when combined with VR2, will make our physical reality undistinguishable from the virtual one. At that point, it will be easy to fully blend interactive film and video game experiences and maximize the immersion. We surely have a lot to look forward to (and also to be afraid of3).

Player: The Vehicle of Control

Another key point of immersion is creating the player’s identity. Control in an interactive film must be limited and channeled into an avatar, a character, or some other level of identity—not just to embody a person but to make it serve as a vehicle for the action. It can be either first-person or third-person style.

Sure, it would be sweet to have control over an entire movie reality, where your real-time interactions with objects shape the unfolding story. You could throw a banana skin into the scene to make someone slip, watching the consequences play out. But who would you be in this scenario? Your role and identity might blur, leaving the experience feeling less like a story and more like an experiment in control.

The most common design for an interactive film is simply trying to convince you that you embody the main character and control it through a limited series of predefined points on the timeline. You role-play and take over the decisions. As in a video game, there are moments where you don’t control the character, and it somehow exercises free will with the story’s natural flow. That’s when your hero says something you’d never say, takes a shot of vodka you’d never drink, or shoots people you don’t even know.

Let’s make you a hero then. Everybody likes heroes!

In Interande, you occasionally take first-person control of Petr, a 2011 dating master wannabe from the Czech Republic. You two share dating weaponry and embark on a quest to convince Dana that you are the future in her simulated life.

The Immersion Quadrant

We already know you must embody the player into an avatar in an immersive way that provides enough control. After all, it’s all about entertainment, but to determine the right level of decision-making granularity, you balance many variable elements.

You must consider the technical capabilities, filmmaking skills, screenplay consistency, movie player capabilities, and depth and complexity of the experience you want to provide. This brings limitations. To get the most out of your interactive film, you must discretize the interactivity to a frame level.

We were building Interande as a prototype to prove the concept and validate the overall potential, so we couldn’t bite off more than we could chew back in 2011. I have always thought of a perfectly rendered game that smoothly switches between movie footage and controllable game back and forth in real-time without the player noticing the difference, but we were nowhere near that far. We had to use the classic way of filming the entire movie material, post-processing it, gamifying it, and then gluing it all together while sacrificing a lot of control potential and smoothness to set up the strings for the player to pull.

Our concept was interactivity on this level:

It’s nothing spectacular, especially not today. However, we did not foresee the many challenges upfront, even when we gave the player control of just a single character on a 15-minute journey. One of the pieces turned out to be the special video player we had to develop to power the movie. Don’t forget to subscribe if you’re curious about how it’s built and what challenges we had to overcome.

Subscribe now

Thank you for reading!

It ran smoothly for over 8 years before it aged and shut down. However, a bit of my digital necromancy leading into its recent resurrection inspired me to also share some details about the story, the technology, memories from behind the scenes, and the reasons it was initially discontinued, only to come back to life and entertain thousands of people every month.

It remains one of those experiences that warms my heart whenever I recall it. The enthusiasm we shared, the naivety, the excellence, the hype, and above all, the very talented people willing to fully dedicate themselves to pulling the rope together.

We wanted to explore the potential, believing that technology had finally advanced to a point where it could successfully awaken the next generation of the interactive film medium, lying in wait.

If you're interested in getting a preview before diving into the full reading, you can find it here. It’s originally available in the Czech language, but there’s also an English voiceover, made exclusively for the iOS version of the app and just recently added to the web version.

Interactive?

The concept of an interactive film was already established when we started. There were many attempts to make a film where the audience is given the ability to affect the narrative through some kind of voting to control the choices of the main heroes. Long before it became possible to produce a movie in an individualized way, where players have their private session to play within, one of the earliest examples I'm aware of, originating from the Czech Republic, was the Kinoautomat (1967).

The projection was exclusive to cinemas and was based on the concept of audience voting between two options at specific moments of the movie.

The doorbell rings. Mr. Novák opens the door, and there’s the neighbor, who is locked out of her apartment, almost naked and wearing only a towel. She begs him to let her in. But his wife might return at any moment, and it's her birthday today. Let the neighbor in or not? …CUT!

When a choice had to be made, the movie was paused, and the host stepped in front of the audience, explaining and narrating the choices. The audience then voted using a controller embedded in their seats, which had two buttons (green and red) for the options. However, it was all a façade. While interactive, the storyline remained shallow and linear as it merged into the same node of the story in the very next scene.

This approach was also utilized in later attempts, where viewers could use telephones or other means to vote on decisions in a movie on their TV. An intermezzo debate was introduced during these moments to allow viewers time to cast their votes. While this concept is democratic, its potential is limited, and most of the latest interactive movies focus solely on a single-player experience. Notable movies I’d like to mention are My One Demand (2015) and Black Mirror: BanderSnatch (2018), which I really like because the main character is a game developer.

Imagine a movie that offers an immersive experience where you, as the viewer, can slow down time, enter the movie, and make a decision through one of the characters to change its course. Intriguing, isn’t it? Now, how can it be made real enough to truly immerse you?

Subscribe for the next article to learn how balancing control and realism is key to the success of interactive film.

Subscribe now

I was recently asked to deliver a couple of talks about AI. I crafted two distinct yet interconnected decks tailored to the audience and the occasion. As both were somehow similar and somehow different at the same time, I originally wanted to share a narrated blend of both, published as a sequence of articles. Instead, I leveraged the original idea and split it into two articles— one based on the lecture’s commented transcript:

Another one based on a few extra slides that weren’t part of the original lecture:

But in the next session, I wanted it to be next level! The topic was up to me, and sometimes, the lack of specification opens up an opportunity for creative expression.

I decided to start by examining how humans model the reality around them to build the session’s skeleton.

Both talks are designed as immersive experiences for the audience rather than mere presentations. The aesthetic blends Matrix-inspired hues with the Courier New font, a staple in the world of computer programming. To blur the lines between human and machine perspectives even more, I intentionally employ hashtags instead of traditional headings.

But this article should be dedicated to the later version of the deck, where I discuss more topics and go into greater depth.

The graphics are created using Midjourney to add an additional layer to the content, striving to complement it by eliciting specific emotions and associations on each slide. As the talk mentions the art of accurately encoding problems and decoding solutions, finding the right imagery through prompting was very important to me, especially when contrasting with the potential of AI's capabilities.

No abstract, instead, the audience received a cognitive shock to engage them:

I chose a list of hashtags with relevance dictated by font size and order. They subtly transition to black, not meant for detailed reading but more to be perceived as a visual imprint on the audience’s retinas. They linger for a moment before we surge into the first, for a pretty deep topic, the Truth…

Subscribe now