#BuildInPublic Manisfesto: Turning Dormant Public Domain into Conversation with PUB (Public User Bot) -
I didn’t start thinking about PUB because I wanted to build another AI product.
I started thinking about it because of silence.
Not the poetic kind—but the vast, institutional silence of information that exists everywhere and yet speaks to so few.
Every day, the internet accumulates more public domain material than any human lifetime could absorb: laws that are rarely reread once passed, archives digitized and quietly forgotten, market filings skimmed only in moments of crisis, technical documentation written with care and then abandoned to search results. These documents are available, technically. They are public. But they are not alive.
The Quiet Life of Public Information
The truth is that in our fast-paced era, there is no time left for the kind of detailed, manual engagement these materials demand. We like to believe we still operate on thorough research and deep reading, but in reality we rely on speed—on summaries, highlights, fragments. Sometimes we verify. Sometimes we don’t. More often than not, we simply can’t afford to reread hundreds of pages to uncover patterns or identify root issues.
We often describe our era as one of radical transparency. Public access is taken as a given, even as it remains fragile. But access alone does not guarantee usability.
Public domain information is scattered across institutions, formats, languages, and decades. It presumes expertise. It presumes time. It presumes that the reader already knows what matters and why. In practice, this turns openness into a technicality rather than an experience.
The result is a quiet contradiction: the most openly available information in history is also among the least engaged with. Not because it lacks value, but because the cost of entry—cognitive, temporal, procedural—has grown too high.
Public domain knowledge, in other words, is available but dormant.
As the founder of @docanalyzer, I spent years focused on private documents—contracts, reports, internal files—and on what happens when people are allowed to interact with them conversationally. The shift was immediate and unmistakable. A document stopped being a static reference and began to function as a thinking partner. People no longer searched. They asked. They challenged. They refined.
The Irritation That Became PUB
PUB did not begin as a polished vision. It began as an irritation.
Why does public information still behave like an artifact rather than a participant? Why does “public” still imply static?
Why can I interrogate my own files but not the laws that govern me?
The answer felt obvious once the question was asked.
What if public information wherever we find it came with the ability to be spoken with?
With PUB, books, records, laws, archives, news, and technical documents become conversational. You don’t adapt to the document. The document adapts to you.
That inversion changes everything.
From Archives to Conversations
Conversation is deeply human. It’s one of our learning practices. How we test ideas. How we navigate complexity.
By allowing users to chat with public information, PUB doesn’t simplify reality—it makes complexity navigable. You can ask naive questions without embarrassment. You can go deep without committing hours. You can extract insight without stripping away context.
This matters especially for professionals—researchers, analysts, writers, creatives, policymakers—people whose work depends not on raw data, but on understanding.
Where I Stand
I didn’t grow up believing that information should be gatekept by format, jargon, or institutional inertia. Building @docanalyzer reinforced that belief. Intelligence isn’t about possessing documents—it’s about interacting with them meaningfully.
PUB is a continuation of that philosophy, applied outward rather than inward. It treats public knowledge not as a monument, but as a living system.
In that sense, PUB is an argument:
That public information deserves a second life.
That access should include dialogue.
That knowledge, once activated, stops being abstract and starts being useful.
PUB has just been launched. It’s currently in its first beta: it’s not perfect—but alive with its long journey of refinement underway.
We chose not to wait for scale, funding, or polish. We chose to begin where public knowledge belongs—in use. In conversation. In the hands of people willing to use and question it.
PUB is an invitation—to read differently, to question more freely, and to participate in shaping how public knowledge lives again. There you can find all sorts of public documents in sections and under categories. They were always yours - you just didn’t know you could really own them.
Perhaps, in giving public information the ability to speak again, we return public to what it was always meant to be.
Sam Altman Wants a Better AI Future for Us.
But How Serious Is He?
It’s been a few weeks since OpenAI released its policy about the future of AI which keeps the people first. Reading it back then it immediately made me think of the blog I wrote about some of these job market shifts back in 2025, in How IT Built the AI That Took Its Jobs. Naturally, I had to write my take on this new document, which took me quite a while.
As an AI entrepreneur who has watched enough technology cycles, I, unfortunately, recognize the pattern: visionary frameworks that diagnose real problems but gloss over hard tradeoffs.
Read on my Substack...
From Coder to Architect
The Age of Nudge Programming
Today, increasingly, we programmers do not write code, we describe outcomes. The direction of effort has shifted from syntax to intention. Instead of spending hours writing code, I prompt something like: Build me a scalable API that handles X, Y, and Z. Instead of debugging line 143, you say: Something’s wrong with how this handles concurrency. Here is the log: <...>
The machine eagerly executes right away, if anything it is too eager, to an extent that sometimes I have to tone my prompts down for it not to overreact. Moreover, it doesn’t then suggest or complete my thought, it replaces it.
I am less and less often type the code myself. And, crucially, I really no longer need to.
This is the new norm, not yet universal, but quite close to being it. There remain holdouts - some engineers who still write every bracket, who distrust the confidence of generated code, who even decline suggestions altogether. There is something almost both strange and artisanal about them now, like watchmakers in an age of quartz. But in every major technology company, in every startup chasing velocity, AI has already been absorbed into the bloodstream of production. To ignore it is not principled; it is uncompetitive.
Because the machine is simply faster. Not marginally faster but categorically so. It does not tire, and if you are checking on them, it won’t stray away from the context. Against that, manual coding begins to look less like mastery and more like resistance.
And yet, something interesting has happened in the displacement.
If the act of writing code has diminished, the act of guiding it has become central. Programming has not vanished; it has metastasized into something softer, stranger, and arguably more cognitive. You watch the system think, you direct it and make sure it didn’t forget the shape of the codebase it read minutes ago. You observe its intermediate steps, its confident missteps. You develop a feel—a kind of tacit literacy—for when it is going wrong. And then you intervene. Not by fixing the code directly, but by rephrasing the instruction.
First you notice. You stop the process. You adjust the prompt. You nudge.
This is “nudge programming”: a discipline defined less by construction than by steering. The programmer becomes a supervisor of trajectories rather than a builder of artifacts. You are not laying bricks; you are redirecting the bulldozer mid-motion.
There is, in this, a subtle but profound shift in authorship. The finished system is still, in some sense, yours—you specified the constraints, validated the outputs, curated the direction. But the path from idea to implementation is no longer legible in the old way. You cannot point to a line and say: I wrote this. Instead, you might say: I caused this to exist.
There is a curious story around this case - Chardet dispute dispute which captures a new fault line in software: if AI can “clean-room” rewrite an entire codebase, developers may claim the right to change its license—while critics argue prior exposure and model training make such independence legally dubious. What once hinged on direct code copying now dissolves into statistical similarity and intent, leaving courts to decide whether AI-assisted rewrites are genuinely new works or derivative in disguise. As the cost of reimplementation collapses, both open-source copyleft and proprietary licensing face a future where code can be replicated, re-licensed, and contested with unprecedented ease.
It is a weaker claim, and a more abstract one. It leaves a lot of ambiguity in between and doesn’t allow to foster ownership. Naturally, this also raises the anxious question: are programmers disappearing?
The answer, I think, might depend on what one believes a programmer is. If programming is the manual inscription of logic in a formal language, then yes—the role is already eroding. In a few years, it may feel as antiquated as writing assembly code for most applications today.
But if programming is the architecture and design of systems, the decomposition of problems, the articulation of intent under constraint—then it is not disappearing at all. It is expanding, even as it becomes less visible. The skill is migrating upward, away from syntax and toward structure.
In this sense, the modern programmer begins to resemble an architect. Not because they draw neat diagrams, but because they operate at the level of systems and trade-offs, delegating execution to increasingly capable agents. The question is no longer How do I implement this? but What should exist, and how should it behave?
Still, there is nostalgia.
There was a tactile satisfaction in writing code by hand, in knowing exactly why something worked, in tracing the flow of logic through your own decisions. That intimacy is harder to find when the machine fills in the gaps faster than you can perceive them. Watching an AI generate a thousand lines in seconds can feel less like creation and more like curation. Perhaps nostalgia exists for some unexplained purpose helpful to evolution in the long run but overall, in technology, it is rarely a viable strategy.
The uncomfortable truth is that manual coding is no longer competitive for many tasks. Not because humans have become worse, but because the baseline has shifted. The machine has absorbed the mechanical aspects of the craft, leaving behind something more ambiguous: judgment, taste, direction. Perhaps in a few years from now students will stop learning coding altogether and concentrate on a new kind of programming - reimagining it from scratch.
The programmer still remains—but in altered form.
Less a writer of code than a conductor of processes. Less a craftsman of syntax than a negotiator of intent. Someone who does not build the system directly, but who knows how to steer it when it begins to drift.
Which, perhaps, was always the deeper skill all along.
PUB update: replaced ad-hoc tags with a 4-dimension taxonomy (form · subject · genre · use-case), redesigned dataset pages to actually tell you what you're looking at, and gave the about page some backbone. #BuildInPublic
The Art of Forgetting
AI, Bun, and the Future of Coding
There is something unexpectedly telling in the idea of @AnthropicAI acquiring Bun. On the surface, it reads like a straightforward business move. But the moment I sat with it, it began to feel like something else—a suggestion that our tools and environments may need to evolve and adapt to the era of AI, if we want to code effectively with it.
For more than a decade, Node.js has been the gravitational center of JavaScript runtime culture. Created in the early 2010s, it earned its dominance honestly—by being early, fast, and flexible enough to absorb a rapidly expanding ecosystem. But long success has a way of aging systems. Node.js today is undeniably powerful, yet deeply layered with history: backward compatibility, legacy decisions, and well-intentioned patches accumulated over time. It works, but it carries memory. And in software, memory often translates into complexity—something AI struggles with more than we like to admit.
Bun enters this picture as a deliberate act of forgetting. Built from scratch, it is not burdened by the obligation to preserve every decision made over the last fifteen years. Instead, it selectively inherits best practices while discarding outdated design choices. The result is a runtime that is cleaner, faster, and—most importantly—tighter. For AI systems, this difference is not cosmetic. It is structural.
Large language models do not experience software history the way humans do. They ingest it all at once. Old syntax, deprecated patterns, half-abandoned conventions—all coexist in the same statistical soup. When an AI generates Node.js code, it may reach for something that once worked, then almost worked, and was eventually replaced without ceremony. The code can be theoretically valid, practically obsolete, and subtly wrong in context—all at the same time.
Node.js, in this sense, offers a clear illustration of a deeper problem in AI-assisted coding. Its richness is inseparable from its anachronisms. Bun, by contrast, presents a narrower and more contemporary surface area of, hopefully, best practices and fewer historical forks, fewer semantic ghosts. It is easier to reason about precisely because there is less to remember. And it is always easier to pull meaning from the surface than from the depths. A cleaner runtime reduces ambiguity. A simpler system lowers the risk of hallucinated solutions drawn from outdated corners of training data.
For me, this represents a subtle but important shift in how we think about “coding with AI.” Until now, the dominant narrative has been about layering intelligence on top of existing workflows—Copilots hovering over editors, models translating intent into syntax. But there is another approach too: reshaping the world so that intelligence has less friction to fight against. Instead of only asking AI to navigate decades of accumulated technical debt, we redesign the terrain it moves through. This is not a revolution, but this helps a lot.
So, in this context, Bun is not merely a faster Node competitor. It is a bet on freshness. I have not yet used it extensively, but I am already curious about how this acquisition will translate into real-world results. The idea is compelling: that starting over—carefully, selectively, without nostalgia—can be an advantage in an AI-first era. It acknowledges something developers have long sensed: the more historical a system becomes, the harder it is for both humans and machines to think clearly within it.
Anthropic’s broader signal here seems to be that adapting to the AI era may require simplification rather than just expansion—making systems more legible, more graspable, more on the surface. But if clarity becomes the new competitive advantage, are we redefining coding for humans, for AI, or for both?
My 2026 idea: making every public domain document chat-able with AI.
Old books, laws, research papers, historical texts → instant conversational AI for anyone.
Starting today & building in public: progress, code, fails, wins. Will make the manifesto tomorrow...
#BuildInPublic
#BuildInPublic Manisfesto: Turning Dormant Public Domain into Conversation with PUB (Public User Bot) -
I didn’t start thinking about PUB because I wanted to build another AI product.
I started thinking about it because of silence.
Not the poetic kind—but the vast, institutional silence of information that exists everywhere and yet speaks to so few.
Every day, the internet accumulates more public domain material than any human lifetime could absorb: laws that are rarely reread once passed, archives digitized and quietly forgotten, market filings skimmed only in moments of crisis, technical documentation written with care and then abandoned to search results. These documents are available, technically. They are public. But they are not alive.
The Quiet Life of Public Information
The truth is that in our fast-paced era, there is no time left for the kind of detailed, manual engagement these materials demand. We like to believe we still operate on thorough research and deep reading, but in reality we rely on speed—on summaries, highlights, fragments. Sometimes we verify. Sometimes we don’t. More often than not, we simply can’t afford to reread hundreds of pages to uncover patterns or identify root issues.
We often describe our era as one of radical transparency. Public access is taken as a given, even as it remains fragile. But access alone does not guarantee usability.
Public domain information is scattered across institutions, formats, languages, and decades. It presumes expertise. It presumes time. It presumes that the reader already knows what matters and why. In practice, this turns openness into a technicality rather than an experience.
The result is a quiet contradiction: the most openly available information in history is also among the least engaged with. Not because it lacks value, but because the cost of entry—cognitive, temporal, procedural—has grown too high.
Public domain knowledge, in other words, is available but dormant.
As the founder of @docanalyzer, I spent years focused on private documents—contracts, reports, internal files—and on what happens when people are allowed to interact with them conversationally. The shift was immediate and unmistakable. A document stopped being a static reference and began to function as a thinking partner. People no longer searched. They asked. They challenged. They refined.
The Irritation That Became PUB
PUB did not begin as a polished vision. It began as an irritation.
Why does public information still behave like an artifact rather than a participant? Why does “public” still imply static?
Why can I interrogate my own files but not the laws that govern me?
The answer felt obvious once the question was asked.
What if public information wherever we find it came with the ability to be spoken with?
With PUB, books, records, laws, archives, news, and technical documents become conversational. You don’t adapt to the document. The document adapts to you.
That inversion changes everything.
From Archives to Conversations
Conversation is deeply human. It’s one of our learning practices. How we test ideas. How we navigate complexity.
By allowing users to chat with public information, PUB doesn’t simplify reality—it makes complexity navigable. You can ask naive questions without embarrassment. You can go deep without committing hours. You can extract insight without stripping away context.
This matters especially for professionals—researchers, analysts, writers, creatives, policymakers—people whose work depends not on raw data, but on understanding.
Where I Stand
I didn’t grow up believing that information should be gatekept by format, jargon, or institutional inertia. Building @docanalyzer reinforced that belief. Intelligence isn’t about possessing documents—it’s about interacting with them meaningfully.
PUB is a continuation of that philosophy, applied outward rather than inward. It treats public knowledge not as a monument, but as a living system.
In that sense, PUB is an argument:
That public information deserves a second life.
That access should include dialogue.
That knowledge, once activated, stops being abstract and starts being useful.
PUB has just been launched. It’s currently in its first beta: it’s not perfect—but alive with its long journey of refinement underway.
We chose not to wait for scale, funding, or polish. We chose to begin where public knowledge belongs—in use. In conversation. In the hands of people willing to use and question it.
PUB is an invitation—to read differently, to question more freely, and to participate in shaping how public knowledge lives again. There you can find all sorts of public documents in sections and under categories. They were always yours - you just didn’t know you could really own them.
Perhaps, in giving public information the ability to speak again, we return public to what it was always meant to be.
Just published a first draft of PUB (Public User Bot) See https://t.co/3XF5kNVmXu - this draft version is using a few books from Project Gutenberg #BuildInPublic
#BuildInPublic Manisfesto: Turning Dormant Public Domain into Conversation with PUB (Public User Bot) -
I didn’t start thinking about PUB because I wanted to build another AI product.
I started thinking about it because of silence.
Not the poetic kind—but the vast, institutional silence of information that exists everywhere and yet speaks to so few.
Every day, the internet accumulates more public domain material than any human lifetime could absorb: laws that are rarely reread once passed, archives digitized and quietly forgotten, market filings skimmed only in moments of crisis, technical documentation written with care and then abandoned to search results. These documents are available, technically. They are public. But they are not alive.
The Quiet Life of Public Information
The truth is that in our fast-paced era, there is no time left for the kind of detailed, manual engagement these materials demand. We like to believe we still operate on thorough research and deep reading, but in reality we rely on speed—on summaries, highlights, fragments. Sometimes we verify. Sometimes we don’t. More often than not, we simply can’t afford to reread hundreds of pages to uncover patterns or identify root issues.
We often describe our era as one of radical transparency. Public access is taken as a given, even as it remains fragile. But access alone does not guarantee usability.
Public domain information is scattered across institutions, formats, languages, and decades. It presumes expertise. It presumes time. It presumes that the reader already knows what matters and why. In practice, this turns openness into a technicality rather than an experience.
The result is a quiet contradiction: the most openly available information in history is also among the least engaged with. Not because it lacks value, but because the cost of entry—cognitive, temporal, procedural—has grown too high.
Public domain knowledge, in other words, is available but dormant.
As the founder of @docanalyzer, I spent years focused on private documents—contracts, reports, internal files—and on what happens when people are allowed to interact with them conversationally. The shift was immediate and unmistakable. A document stopped being a static reference and began to function as a thinking partner. People no longer searched. They asked. They challenged. They refined.
The Irritation That Became PUB
PUB did not begin as a polished vision. It began as an irritation.
Why does public information still behave like an artifact rather than a participant? Why does “public” still imply static?
Why can I interrogate my own files but not the laws that govern me?
The answer felt obvious once the question was asked.
What if public information wherever we find it came with the ability to be spoken with?
With PUB, books, records, laws, archives, news, and technical documents become conversational. You don’t adapt to the document. The document adapts to you.
That inversion changes everything.
From Archives to Conversations
Conversation is deeply human. It’s one of our learning practices. How we test ideas. How we navigate complexity.
By allowing users to chat with public information, PUB doesn’t simplify reality—it makes complexity navigable. You can ask naive questions without embarrassment. You can go deep without committing hours. You can extract insight without stripping away context.
This matters especially for professionals—researchers, analysts, writers, creatives, policymakers—people whose work depends not on raw data, but on understanding.
Where I Stand
I didn’t grow up believing that information should be gatekept by format, jargon, or institutional inertia. Building @docanalyzer reinforced that belief. Intelligence isn’t about possessing documents—it’s about interacting with them meaningfully.
PUB is a continuation of that philosophy, applied outward rather than inward. It treats public knowledge not as a monument, but as a living system.
In that sense, PUB is an argument:
That public information deserves a second life.
That access should include dialogue.
That knowledge, once activated, stops being abstract and starts being useful.
PUB has just been launched. It’s currently in its first beta: it’s not perfect—but alive with its long journey of refinement underway.
We chose not to wait for scale, funding, or polish. We chose to begin where public knowledge belongs—in use. In conversation. In the hands of people willing to use and question it.
PUB is an invitation—to read differently, to question more freely, and to participate in shaping how public knowledge lives again. There you can find all sorts of public documents in sections and under categories. They were always yours - you just didn’t know you could really own them.
Perhaps, in giving public information the ability to speak again, we return public to what it was always meant to be.
Do you remember when you first noticed that two things—treated by everyone as separate—might actually be the same thing in disguise? It takes years, and very particular environments, to develop the context needed for that realization. Often, the problem was never hidden in the data, but embedded in the ideas themselves, waiting to be placed beside a larger pattern.
Then Newton’s breakthrough wasn’t an apple to the head. It was the realization that the force pulling apples downward might be the same principle governing the moon’s orbit—that heaven and earth obeyed one law. Archimedes didn’t just notice water spilling from a tub; he recognized displacement as a way to measure volume indirectly. Fleming’s moldy petri dish mattered only because he looked at bacterial death and thought: this is an antagonistic chemical effect at microscopic scale—what if we could use it?
Read the post on Substack...
Every few weeks, I monitor the AI model leaderboards to check the pulse of the field — and to pressure-test my own intuitions. It feels like looking up Olympics results after watching the games—one week GPT sits comfortably on top, the next Grok surges ahead, then Gemini appears with a decisive jump.
But lately, based on what I have been seeing, something new is happening. The race isn’t just about who has the smartest model anymore. It’s about who can provide quality for the lowest price.
Keep reading on Substack…