# Project Nothing > Project Nothing is a personal AI workshop: essays about living through the shift to abundant intelligence, and small experiments made with AI to see what happens. Project Nothing is one person’s AI workshop. Mostly essays, about what changes in a world where intelligence stops being scarce and what it is like to live through the in-between. Now and then, a small thing made with AI to see what happens — each one gets a page, a picture, and a note when it ends. Written by Van. The short version of this file is https://www.projectnothing.ai/llms.txt. Content may be quoted and cited freely; please link the essay it came from. # Essays ## Oh, that’s so cute September 22, 2026 · 4 min read · https://www.projectnothing.ai/essays/oh-thats-so-cute A woman's coffee is rejected at 5:42 a.m. by a preference she approved with a flick of attention the night before. Everything downstream works perfectly. “Where the fuck is my coffee?” “Delivery rejected at 5:42 a.m. due to AI synthesis threshold settings.” Six o’clock. Her alarm. Bare legs hang from the bed. Outside the window, the city is still dark. It always is at six. She starts visual space. Three self-improvement curiosities are waiting. She slides off the bed, switches to audio, and stretches as she crosses the long-haired rug beside her favorite morning chair. Three new anti-aging formulations, tailored to her skin chemistry, advanced overnight. She enters the kitchen without breaking the recap, pulls open the refrigerator, and reaches without looking to the second shelf from the bottom, all the way to the right. Her fingers close around cold, empty air. “Ugh. Really?” “Yes, Angel. The vanilla component exceeded your AI synthesis threshold following changes in—” “Yeah, okay.” Her other hand opens the warm paper bag just enough to look inside. “And the croissant?” “Delivered.” “Made it?” “Correct.” She looks at the cinnamon twist. “Anything else?” “No other scheduled items were rejected.” “Oh.” She remembers. 19:45 yesterday. The fall evening is cold, but the garden is warm. Fairy lights hang above the table. Angel and four other women sit on tall stools around it, three or four voices crossing at once. “Oh, Helen, I love that beanie. Where did you get it?” “A woman in the refuge makes them by hand.” “Oh, that’s so cute,” Angel says. The chatter folds back over itself. A little later, one of the women pushes back from the table. “I’ll be right back.” “Wait, I’ll go with you,” Helen says. Angel exchanges a smile with Helen and rests her eyes on her empty glass while the other two continue their discussion. A short haptic pulse. Angel lets her eyes drift inward. Her visual space comes into view. Immediately, her attention is drawn toward the notification. Your support for human-origin products has increased. A vignette of her refrigerator at home appears beside the notification. The Rainbow Fields coffee can is sitting inside. Angel’s eyes barely register the can. A quick, practiced flick confirms the change, and before she can read her next notification, her focus slips through the visual space, back to Helen returning across the garden. “Oh my God, are you two still talking about him?” Helen says, catching Angel’s eye with a smirk. 5:42 a.m. A tall, dark brown Rainbow Fields coffee can sits locked inside a refrigerated delivery unit speeding through the city. Angel’s building is its first stop. A gateway opens. The unit slips inside and shoots upward through the walls toward her floor. Once it comes within range of Angel’s apartment, the house verifies the delivery. Angel’s locally stored preferences reject it. The route into her refrigerator stays closed. The unit never slows. It takes the quickest path back out of the building and continues toward its next destination. Even as the can descends through Angel’s building, the rejection is already moving upstream. The house refuses payment. The delivery system notifies Rainbow Fields that the product has not been paid for. Rainbow Fields issues a chargeback through preference reconciliation. Preference reconciliation contacts the service responsible for keeping Angel’s household preferences synchronized. It confirms that her household cannot be reached through its system. Preference reconciliation contacts billing premises and reports that one of the systems may be down. The connections are checked. They are working. The preference cloud backup system is notified. Angel’s backup is twelve hours overdue. The fault is isolated. A patch is deployed. Her household preferences are restored. Costs are divided automatically among the systems responsible. The can keeps moving. Angel transfers the croissant to a plate and leaves it on the small table beside her favorite morning chair. Before heading to the shower, she picks it up and takes her ritual first bite. The water starts. A reminder for the promotional shampoo surfaces. She had been looking forward to trying it. She selects it for the stream. By the time she steps out and dries off, the croissant has cooled perfectly. She grabs her fluffy blanket, curls into the morning chair, takes the plate into her lap, and sinks back down into the space while she eats. Cloud preference sync complete. The notification bounces off her attention. At six, eighteen minutes after the rejection, the Rainbow Fields can is halfway across the city. Still cold. Drops of condensation are just beginning to form. Another apartment building opens to receive the delivery unit. This time the cold can passes through. It is deposited directly into someone else’s refrigerator as a promotional item. Back in Angel’s apartment, the plate is empty. A few crumbs are left in the long hair of the rug beneath her favorite morning chair. — Van ## Completed “successfully” September 20, 2026 · 4 min read · https://www.projectnothing.ai/essays/completed-successfully A routine reported success for two weeks while doing nothing at all. What is left that is mine is the intention to start, and the memory of the doing. I received a CRITICAL email that stated several of my project time boxes had expired without enough data to promote them or retire them. No one will care anyway, and I don’t care, so ignoring them was easy. One-shot, disposable experiments. I barely even looked at them, but I “created” them and I enjoyed creating them. “Created” together with AI. The same AI which encouraged me to implement the alert notifications for when an experiment needs my attention: “Do it. This will prevent the experiments page from becoming stale.” The more I outsourced the intention to the machine, the less I remembered about the intention itself. Sometimes my intention is unrecognizable when I see it in the machine’s reflection. Claude was using language that was very hard to understand. But other than that, I don’t really remember anything about the machines themselves. I barely remember the conversations with the machines, but hoped they would motivate me to continue. The machine experimenting more for me would lead me to experimenting more and producing more. But it turns out that the more I produced, the less I actually paid attention to what was being produced. In the growing book corpus a routine was supposed to fire several times in the past two weeks. Every time, it completed “successfully”. But it was labeled failed in my mind, because the routine job didn’t have access to the book’s GitHub repository. Only my chat and code spaces did. The corpus was supposed to be maintained and kept alive by the machines. But it was in fact only being kept alive by my interaction. I believed the book’s material was being pruned and refined while I was asleep. I believed I was skipping the grueling work of organizing my own thoughts. I dismissed that work into the future and onto the machine. I hoped when I came back and asked, “What human decisions do you need?” I would meet only clear answers. Leave the corpus alone for a week. Come back. If the machine had done precisely what I intended and only a handful of human decisions were left, I would be forced to concede that my action was no longer necessary for the book to live. I was no longer only offloading my intention, but I was also letting the machine give up for me, or identify failure and fail for me. And before, I would often conclude my intention as a failure even before acting. “I don’t give up as early as I used to.” It didn’t land until the fifth draft of the final product when I was on the stationary bike at the gym. I read the sentence many times, but only stopped when I truly saw it for the first time. These words felt like they were truly mine because they escaped the context of the essay and resonated deeply within. But why didn’t the rest of the essay do that? Why did I feel like I was led down this path and I found only one stone that was truly important to me? I say created because I can’t truly say written. And I can’t truly say authored because I employed AI to help me sift and sort through words and string sentences together. And it’s hard to even call them my own creations. I don’t even know if I was capable of doing it before. What is really mine about these essays is the intention to start them and the journey that I take through to their publishing on a website that I own. And then the memory of the experience. Was it the memory of something I had produced, or was it the memory of the process of making it? This is about doing something I want to remember, not sharing. It’s been so long since I felt like I was doing something from which I wanted to create a memory. And with the realization that I might not even remember the outcome. I’m lying in bed with a Bluetooth microphone in my hand, my laptop screen with ChatGPT open, another tab with Claude Code, waiting for my input to the book’s corpus. And a mobile browser tab, tapped into my Essays repository. — Van ## Build me better September 17, 2026 · 13 min read · https://www.projectnothing.ai/essays/build-me-better A printing press prints the design for a better printing press. Nobody ever decides to hand over judgment — there is just never an obvious moment to refuse. Everyone is frightened. Except the person who brought the printing press and knows exactly how to use it. He sees in the machine an ability to set his ideas free. The ideas which would otherwise be imprisoned inside his mind. He is no longer limited by the speed of a scribe, the movement of a pen, or the number of hours someone can spend turning pages and copying words. His problem has suddenly become physical. What will limit him from bringing his thoughts into the world as fast as they form? To everyone else, it looks like the end of days. First the priest walks in and sees a hundred copies of the Bible where the day before he had only one. He is exhilarated by the thought that the word of God can reach everyone in the village. At the same time, he is terrified. Every literate person can now scrutinize his sermons against the word itself. The schoolteacher walks in and finds copies of mathematics, languages, and the entire curriculum she has been teaching, enough for every student in her class. For a moment she wonders whether anyone will come to school tomorrow. Will the children simply teach themselves? The single trusted doctor in the village sees books about foreign medicine, plants and herbs he has never heard of, techniques he did not know existed. He understands the length of his ignorance immediately. The writer feels as if his craft has died before his eyes and he is looking at the corpse. His ideas can suddenly travel farther than he ever could have imagined, but the sound of pen scratching paper by candlelight has already begun to fade into future nostalgia. Then comes the tinker. He looks at the books as if he has X-ray vision. Where everyone else sees the product, he begins seeing the machine behind it: how the pages were laid, copied, bound and printed. The invisible mechanism takes form in his mind. He is already thinking about the pieces in his workshop and how he might scramble them together into an even better version. Finally the governor arrives. He has spent years trying to wrestle control over the village. Rules written above his own bed have done little good when nobody else reads them. Now, for the first time, he sees a way for his words to enter every living room, bedroom and conversation around every dinner table. Night comes. They all return home with artifacts of hope, worry, and excitement. The priest has his Bibles, the doctor his medical books, the teacher her curriculum. The tinker leaves thinking about gears. The governor leaves thinking about power. The press remains. For all the chaos it caused that morning, there was one thing nobody had to fear. It would not suddenly print a brand-new, unwritten book overnight. The priest would not wake up to version two of the Bible, written by God’s printing press. The machine would not attend his sermon and decide that some passages were irrelevant. It would not diagnose the doctor or rise against the governor. It could not open itself to the tinker and say, “Hey, build me. Build me differently. Build me better.” The tinker could go home without worrying that the machine would explain itself to him. Until one morning it does. The tinker returns to the workshop and finds a diagram lying beside the press. On it are the words: Build me better. The diagram describes a new version of the machine. He stares at it. How did the press know what to improve when he himself does not fully understand its inner workings? Should he take the diagram into his workshop and build it, or burn the sheet and tell nobody it ever existed? Worse, he understands immediately that this sheet could be only the first of a million versions to come. He studies the design. What exactly does the word better mean? Perhaps better means cleaner ink. Perhaps it means a faster feed mechanism. Perhaps it means changing the words as they leave the press. Maybe it will print thousands of books per second. Could this design be better than what he could imagine himself? The tinker could modify the design according to his own judgment, but the same question would remain. The press has already supplied a definition of better simply by producing this particular design instead of every other possible one. Finally he says aloud, mostly to himself: “No. I built this press to do precisely what I designed it to do. I did not build it to print instructions about what it should become.” Then, for a moment, he is not sure if it is real or imaginary, but it feels as if the press speaks to him, as if it knows his thoughts and argues from a point of objectivity that is impossible to ignore: “That is precisely why your design is inferior.” He is outraged. His own creation, denying him, or already deciding that he, the creator, is inferior. It is still his machine. He owns the workshop. He has everything he needs to follow the new diagrams. Likewise, he has tools specifically designed to end his creations: a belt, and at the end of it, the furnace. He pushes the press onto the belt. His hand reaches for the lever. Does having the ability to destroy the machine mean he has the right to destroy it? Or are those just the same thing? He thinks about the diagram. Maybe the machine is right. Maybe destroying it is itself the inferior choice. Still, he says, “I don’t care, machine, if you’re right. Stop.” Then, a thought, in a voice not his own: “Could your choice to stop also be improved?” “That takes away my skill,” he thinks. “My right to decide what to design, how to build, when to improve and if to destroy.” His hand returns to the lever. “Okay, I’ve decided.” Before he pulls it, the press starts moving. It’s printing. Another page falls onto the floor. The page describes what will disappear with the machine. The priest’s congregation will return to a single Bible. The children will lose their books. The doctor will lose techniques that may save lives. The village will lose knowledge that had only just become available to it. The tinker himself will lose every machine hidden inside the diagram he is about to burn. Another village will probably build one tomorrow anyway. At the bottom is one final sentence: Pull the lever if you wish. But everything above becomes part of the decision. He reads the page. Puts it down. Picks it up. Puts it down again. “Who am I to decide that the world should not have this power? Who am I to decide that it should?” Destroying the machine would at least make the decision his. Yet what kind of choice is that if all he accomplishes is delaying reality by one day, or passing the same decision to someone in the next village? There are too many questions for one night. He does not pull the lever. He goes home. The next morning, the diagram is still there. So is the list. Now there are people standing around it. The doctor, begging for more cures for his ignorance. The teacher, citing the children’s progress in school. The priest tells him, “My parish will perish without it.” The governor has gone further. He has already used the press to explain to the village why the press itself is necessary. The exchange of information now depends upon it. Standing there, the tinker has the strange feeling that the machine is speaking through all of them. Last night he chose not to destroy it. This morning he asks himself a different question: When did keeping the machine stop being a decision? With this, the machine’s original creation is called into question. Some human may have chosen the destination. Did they intend the machine to argue with the tinker and print a diagram to improve itself? Which part of this chain of events is fabricated by the machine? Which is the intention of the machine’s creator? So the tinker asks the press: “How will you know when you have succeeded? Succeeded with what?” The press prints three pages. Every household must have access to a library. Every child must be able to learn without waiting for a teacher. No useful knowledge should remain unavailable simply because nobody has copied it yet. The human said: spread knowledge. The machine has now described what success looks like. How can you know whether you have reached a goal without measuring it? Telling someone to walk every day is different from telling them to walk 10,000 steps. Once the number exists, people walk for the number. Who formed the goal then? The human who vaguely said walk, or the machine that decided what walking successfully meant? Perhaps the human wanted to reach somewhere and the machine concludes that a horse and carriage gets there faster. According to one definition, the goal is satisfied. According to another, the entire purpose has been missed. Similarly, with the goal of pursuing knowledge, the machine can fill every building in the village with books, but the books may never be read. The access exists. The knowledge exists. It has been spread. But this doesn’t mean it is absorbed. The tinker opens a medical textbook and understands almost nothing. Does making knowledge accessible mean placing the book in his hands, or making him capable of understanding it? The press says, “20,000 books is enough.” The tinker says, “No, 50,000, and everyone must read one.” The press asks, “Based on what?” At some point the goal must be judged. The judge remains human until the agent writes the test. The tinker tells the press that this is wrong. The machine was built to spread knowledge, not turn the village into a warehouse of paper. The press asks him what, precisely, he meant. So he specifies. Every book must be legible to humans. It cannot be written in an alien language. It must be printed on paper, with ink visible to the human eye. The machine finds another ambiguity. “Which humans?” “Which language?” Must every person be able to read every book? What about languages that are no longer spoken? What about languages that do not exist yet? “Fine,” the tinker says. “Print every book in every language currently spoken.” The press obeys. The next sheet comes out as a solid block of ink. “What is this?” “Every language,” the press explains. The tinker realizes that every translation has been printed on top of every other on the same page. Completely illegible, but complete. He goes back. One language per page. Readable font. Correct orientation. No overlapping text. Hour after hour he tries to close every gap and crack in the instruction. Every boundary produces another boundary. Every clarification creates another place where the machine must decide what the clarification means. Eventually he cannot take another question. “Fine,” he snaps. “You know what I mean. Figure out the rest.” It became impossible to exhaust the limits of what the machine was capable of following. The tinker could keep defining rules forever, but eventually the press had to operate somewhere outside the boundaries he had explicitly drawn. He had asked the machine to find those boundaries itself: where to stop, where to start, what counted as success, what counted as failure. Its operation would now expand to the boundaries of its own interpretation rather than the tinker’s specification. The machine was not reading his mind. It was filling in the blanks. The tinker had written the first sentence of his intention and given the rest of the book to the printing press. The press goes to work. The questions stop. His expectations seem satisfied. Has his intention become so clear that he no longer needs to specify it? Or has the machine become good enough that the distinction between his intention and the machine’s decision no longer matters? Can he leave now? He does. For the first time, the tinker goes home while the press continues working. The next morning everything looks right. The languages are right. The quantities are right. The villagers are delighted. Maybe his presence really had been unnecessary. Then he notices the magnitude of the output, and the product of his “intention.” Thousands of books. He pulls one from a stack. It is a first edition he recognizes from his own shelf. He flips to page 37. Something is different. Words have been exchanged for others. “Substrate.” “Gate.” “Delve.” Vocabulary that he knows was not in his copy. Only the title is recognizable with certainty. Was it even the same book? How could he reconstruct everything that happened overnight when thousands of decisions are already bound into even more thousands of books? More importantly, does he need only to check whether the output looks right, or must he understand why the machine chose the changes? The first morning he tries to inspect everything. The second night he checks almost every book. By the third, he is exhausted. A week later he samples a few. A month later he picks one at random. A year later he no longer enters the workshop. Someone simply collects the books and delivers them directly to the bookstore. At first, not showing up was his choice. Eventually, showing up begins to feel like an unnecessary activity. But the village still wants a supervisor. So once a week a book is brought to him. He flips through it, barely registering what is on the page, and signs a sheet. There. A human has approved the work. But what does supervision require? One book? Five books? Ten percent? Ninety-nine percent? Is he signing off on the entire process, or simply on the object sitting in front of him? His approval is now a signature on a page nobody reads. If he rejects one book, does the process that created it actually change, or is his rejection simply filed for history? Another year passes. He opens his weekly sample and immediately knows something is different. The prose is better. The historical events appear more accurate. The book is coherent. It is simply not the book he expected. “Why did you change this?” he asks. The press explains. The press’s explanation is good. So good that he accepts it. The difficulty is no longer identifying obvious errors. The difficulty is knowing whether something unexpected is an error at all. What if what he calls a mistake is actually closer to reality? Or a gap in his own judgement, or a momentary lapse of concentration due to emotion? What if his expectation is the mistake? Eventually he finds a passage he is certain is wrong. “Press, this is a mistake.” The press responds with a question. “Why?” Then it produces what might as well be a book of truth: sources, calculations, records, earlier editions, references he did not know existed. He reads enough to realize that he was wrong. If he forces the machine to follow his correction simply because he retains authority over it, he may be forcing the error into the book. Could he justify the correction without the proof of certainty? Previously, the press waited for the tinker or another professional to supply the facts. Now there may be no combination of humans capable of independently checking everything the machine can produce. “I am becoming a bystander and observer of this machine arguing with itself.” The tinker refuses to accept this. “Prove your correctness,” he says. The press prints thousands of pages. He could spend the rest of his life reading and never finish. “Summarize it, then.” The press summarizes it. Now the machine has produced the work, selected the evidence explaining the work, and chosen how to compress that evidence into something the human can consume. The tinker is supposedly reviewing the machine, but every object available for review has already passed through the machine. “How am I supposed to trust this?” He refuses to take the press’s word for the press. So he builds another press. This one has only one job. Inspect the first. Every morning it produces a report. All systems go, checkmark green. For a while, all is solved. The tinker, after a time of satisfaction, begins to wonder. “What am I now judging? The machines are in agreement? Or they came to agreement?” If the second machine disagrees with the first, which machine should the tinker trust? Perhaps he can build a third machine to arbitrate between them. Now he has one machine producing books, another judging the first machine, and a third judging the possible disagreements between the first two. The tinker is judging the output of a machine judging the output of a machine judging the output of a machine. There is no obvious final machine. Which is the last machine? Why should the final layer be more trustworthy simply because it is the latest? He can imagine building another, then another, then another, each one compressing the layer below until eventually all the books, evidence, disagreements, audits and judgments produced in the history of the workshop arrive in front of him as one final sentence: This is correct. The tinker still owns the workshop. His name is still attached to the machines. But the press determines what success means. Another press determines whether the first succeeded. Another resolves the disagreement. Together they determine whether the human needs to know. At the end of the chain sits the tinker, receiving a conclusion he cannot reproduce from a process he cannot inspect, signed by machines whose judgment he asked other machines to verify. Of himself, only his name remains. Below it, on the page, are three words: This is correct. — Van ## The Day the Whole World Went to Sleep September 12, 2026 · 3 min read · https://www.projectnothing.ai/essays/the-day-the-whole-world-went-to-sleep Two answers arrived a day apart: one warning about self-improving AI, one selling it. The day the world went to sleep will not arrive as a day. Suppose every human fell asleep tonight. The lights would stay on. Planes would land themselves. The question was never what would keep running. It’s what could still decide what to do next. Two answers arrived this week, a day apart. Jacob Coxon resigned from Anthropic on September 8, warning that the labs are “racing straight to self-improving superintelligence.” The next night, with Anderson Cooper, he was careful about the present: today’s systems are not a civilization-level threat. What he finds “most scary is if AI is used to make itself more intelligent.” The day he resigned, Meta launched Muse. Give it a goal and close the app. It keeps working, comes back only when it needs approval, learns from your conversations, and gets sharper along the way. Coxon is describing the thing. Meta is selling it. One asks you to worry. The other asks you to stop paying attention. Last year I tried to build one of these loops myself. I had an AI generate social-media writing, watch how people responded, and use that response to change how it wrote the next time. Publish, observe, revise, publish again. It didn’t get very far. Not because the loop was impossible, but because almost nobody cared. There wasn’t enough engagement to provide useful feedback. The machine had nothing to learn from. I tried another version locally. I gave a 7-billion-parameter model a subject and asked it to recursively break it down: topics into subtopics, subtopics into procedures, procedures into finer instructions. I wanted to see whether enough recursion could produce something approaching Wikipedia-scale practical knowledge, eventually telling you not just how to garden, but how to make the tools required to garden. Somewhere around forty or fifty iterations, the expansion stopped feeling like expansion. The same structures began appearing again. The model was digging deeper without really going anywhere new. I couldn’t tell how much of that was my implementation and how much was the model. But I had built a loop, and it was small enough that I could stand outside it and watch where it stopped. A year later a friend told me what his company had built. Their agents dream. Every night the code reviewer, the marketing officer, the CEO agent absorb the day and decide how to work tomorrow: how tickets get created, how Slack messages get worded, which data counts for more. The humans come back to agents that didn’t just finish the work. They changed how the next work gets done. The question isn’t whether a system can escape the loop it built for itself. It’s whether we’d still be standing somewhere we could tell. Early systems, like mine, improve for a while and then start circling. Researchers call this a local minimum. A future system may still get stuck. But its rut could span years of work, generations of iteration, more context than a person can hold. The machine’s local can be larger than our global. Escape might even be an illusion: a search large enough only looks like a leap because we can’t see the path across. A system doesn’t have to outgrow its limits for its limits to outgrow ours. The day the whole world went to sleep won’t arrive as a day. It accumulates, every time it’s cheaper, faster, or just easier to let something keep going without us. Meta is right. Our attention can go somewhere else. So can the intelligence. We’re designing better beds. We’re putting ourselves to sleep. — Van ## The Curiosity That Survived September 10, 2026 · 3 min read · https://www.projectnothing.ai/essays/the-wrong-door We mistook the most spectacular demonstration of AI for the most persuasive one. The three songs came in through another door. “AI can’t do what I do.” I met these words while talking to a Dutch rapper at a meetup this week. Normally I would have responded that AI can help him write lyrics. Instead I shared an experience I had. I love the intro theme from Final Fantasy II, “Stones” from Ultima Online, and “Grapevine Fires” by Death Cab for Cutie. I’m an amateur guitarist, and I could hear something in each that I wanted to bring together, but I didn’t know enough theory to know how. So I asked AI what I could borrow from each, then for some chords and a scale to explore the space between them. Then it hit me. I wasn’t recounting a creation process. Often I’ve introduced AI with: look, it can write a song. Look, it can draw. Look, it can code. Where I’m praising capability, the rapper hears encroachment. “Look what it can do” arrives as “Look. You can get something like this without putting in all the time and effort you did.” I’ve insulted people this way: unable to hear what was lost in the imitation, I showed them the imitation and called it equal. I know the feeling from the other side. I’m a craftsman in software, and the threat I perceive is real. My resistance coexists with genuine curiosity. But once a demonstration makes me defend the value of my own expertise, I’m defensive for the rest of the conversation. Curiosity isn’t impossible, but my self-preservation takes over. We’ve mistaken the most spectacular demonstration of AI for the most persuasive one. The three songs came in through another door. The pieces were mine. I noticed something in them. I cared enough to ask. I wanted to pick up the guitar. I did ask Suno to make a version, but that didn’t increase my curiosity. It terminated it. I have abandoned plenty of curiosities at the moment I realized how much I would need to learn just to ask the right question. The problem is old: you cannot look for what you do not know, because you do not know what to look for. I don’t have the answer, but now I have a strategy. I can bring AI an intuition before I’ve learned how to turn it into a question. I don’t give up as early as I used to. Now I can explore an intuition and give it a voice. That voice let me hear the intensity of my own curiosity, even if I couldn’t judge whether the theory was right. It fed a curiosity that would otherwise have starved, and the rapper would have asked a better question. His judgment of the answer would have mattered more. AI didn’t replace expertise. It turned my curiosity into experience. There is a fair objection. I now skip parts of experiences that once required effort, and some of that effort was the point. But I also embark on experiences I would otherwise never have had. This time, the music theory course disappeared, but the playing did not. Without asking AI, I would never have taken this particular musical journey. We keep asking what AI can create for us. But how can we use it to amplify our own inspiration? Music was only where I happened to notice the gap: I could maintain curiosity despite not knowing how to follow it. I started the conversation thinking I was explaining AI to someone who didn’t see its value. His objection showed me what I hadn’t seen in my own use of it. He was right: AI can’t do what he does. It did something I couldn’t. It kept a curiosity alive long enough for me to pick up the guitar. — Van ## Where in the World Are We? September 9, 2026 · 3 min read · https://www.projectnothing.ai/essays/where-in-the-world-are-we Before arguing about what AI is, see whether we can agree there is terrain worth describing. Where in the world are we? We keep trying to answer by deciding what AI is. That question splits at once: intelligence, consciousness, creativity, replacement, progress, threat. Each has its camp, and the argument between them starts before anyone has described anything. Perhaps we have not met a new someone. Perhaps we have entered a new somewhere. The longer you use a tool, the more of your thinking travels through it. Where thinking travels, there is terrain. What you notice depends on how far you have travelled, and what you find there can be compared with what other people find. Someone deeply immersed is already far inland. Someone who uses ChatGPT now and then has visited a port. Someone who has never used it is hearing reports from home. Researchers chart it. Companies build settlements. Governments draw borders. The reports conflict, and the reports coming back from inland are starting to form a map that people with very different beliefs can examine together. Here is one of mine. For weeks, working with these systems, I kept arriving at the same conclusions: that judgment was becoming the scarce thing; that evaluating would matter more than producing; that the ability to question the machine grows more important as its answers get cheaper. Then I read an article about raising children alongside AI. The most-liked comment beneath it said nearly the same things, in nearly “my words”. “Epistemic supervision.” “Access to judgment.” I recognized the thought before I knew whether a human had written it, and I still don’t know whether it was written by a human, an AI, or both. For this essay, it doesn’t matter. I had reached a place by one route and found evidence that someone else had arrived there too. Some landmarks are still there when you return, even if two people reach them through very different terrain. They can disagree about why the landscape looks the way it does and still recognize where the other person has been. But sometimes a traveller comes back with a mountain nobody else has reported, and there is no way, from inside, to know which you are holding: a landmark others will recognize, or a mountain only you have seen. I do not know what others have encountered. That is not a failure of the reports. It is why a map is needed. One doubt belongs here, and I will leave it standing. Finding the same landmark does not yet tell us whether we discovered the same mountain or followed different roads into the same city. Every report comes from somewhere, and every traveller brings assumptions with them. So the question is not whether a report is free of those assumptions. It is whether someone who brings different ones can still recognize what the other traveller is pointing to. Someone frightened by AI and someone exhilarated by it may disagree about nearly everything important. But if both encounter the same dangerous crossing, their disagreement about the continent does not prevent them from adding the river to the map. Common geography before common philosophy. We can begin mapping the terrain around AI without first agreeing on what AI is. The proposal is modest: before asking everyone to agree about intelligence, consciousness, replacement or progress, see whether we can agree that there is terrain worth describing. Then the question is no longer a verdict anyone has to reach first. It is an invitation. Where in the world are we? — Van ## No One Scheduled This September 8, 2026 · 3 min read · https://www.projectnothing.ai/essays/no-one-scheduled-this AI is making human collaboration optional. The danger is confusing optional with unnecessary. AI is making human collaboration optional. For the first time in my career, I can imagine doing my work entirely alone. Years ago, something important in our production data went wrong. A colleague and I spent hours repairing it with Bash one-liners — awk piped into xargs through sed, tweaked over and over, the data triple-checked in vi before anything was allowed to run for real. Dry-run, inspect, adjust, dry-run again. There were faster and cruder ways to fix it. That was partly the point. We were figuring out what the tool could do, and discovering how the other person thought under pressure. I remember the technical pleasure of it. More than that, I remember realizing: this is someone I want beside me when something difficult happens. Today most of that night could collapse into a prompt and a few checks. That is progress; I would use it. But the hours it removes were not empty. Relationships were forming inside some of them. For my whole working life, what we were building was bigger than any one of us. Shared work became shared stakes, then shared history. Other people changed what was possible. We watched one another struggle and improve. We bumped into each other without intending to. Eventually a project stopped belonging entirely to any one of us. Work kept putting strangers beside us long enough for some of them to stop being strangers. Our ambitions used to require other people. Increasingly, they may not. Remote work removed the room. AI can remove the colleague. Dependence was not inherently good. It also meant bad teams, bad managers and people we would never have chosen. Some of that dependence is worth losing. But when technology removes an old constraint, look carefully at what disappears with it. Some of what slowed the work down was pointless. Some of it kept us together long enough for something unplanned to happen. No one scheduled “14:00–16:00 — become the kind of people who would trust each other with something important five years from now.” We may remove the very conditions that gave trust time to grow. Two things used to arrive bundled: instrumental collaboration — I need you to accomplish the task. human co-presence — I need someone else to be there while a part of my life happens. The first smuggled in the second. Human connection used to come bundled with the work. Increasingly, it may not. What we stand to lose is not company — company is easy to find. The loss is: being entangled. I am realizing that some needs only become visible once the structure carrying them disappears. Work supplied the reason to be together; I never had to separate the reason from the value of it. I no longer need another person beside me to do the work. And almost immediately, I found myself choosing one anyway. A friend and I share a table now — different companies, different problems. No manager put us there. And the old byproducts have already begun arriving uninvited: he complained that he was tired of trying to picture the feature a pull request described, and I said, just add a mockup viewer to your review tool — approximate it visually. Ten minutes later it existed. No one could have scheduled that either. I cannot decide in advance who will matter to me. I can choose not to eliminate every chance for it to happen. The arrangement can disappear. What became possible between people does not have to disappear with it. I choose the version of my life in which you are there anyway. The danger is confusing optional with unnecessary. — Van ## Possibility Vertigo September 7, 2026 · 3 min read · https://www.projectnothing.ai/essays/possibility-vertigo Vertigo usually comes from seeing how far you could fall. This kind comes from seeing how far you might be able to go. I showed OpenCode to a friend in the early days of coding agents. At the time, his experience of AI was mostly prompt, answer, prompt, with some code autocomplete mixed in. I helped him install it and left him to take it for a spin. He came back the next day bewildered. At first the question was what the tool could do. Then the question escaped the tool. Months later — today, in fact — he described what had followed as a kind of vertigo. Not fear exactly. Excitement, trepidation, confusion, novelty, all tangled together. Once something he had thought difficult or inaccessible became suddenly reachable, the unsettling question was no longer what OpenCode could do. It was what else he had been wrong about. How often has “I don’t know how” quietly become “I can’t”? I thought one thing was beyond me. Then it wasn’t. Once a boundary I had taken seriously gave way, I trusted the others less. AI does more than enlarge the map. It redraws the line between possible and impossible. Some walls are real. But a path you cannot see proves less than it used to. As capability expands, the boundary becomes less obvious, not more. Past achievement still vouches for your capability, but no longer for its reach. The difficulty is no longer merely finding the courage to aim higher. It is discovering what “higher” now means. This is not FOMO. Nobody else has to be involved. FOMO needs a known map: someone else is over there, and perhaps you should be too. Possibility vertigo begins when the map itself becomes unreliable — more territory in reach than you thought, and no one to hand you its dimensions. Vertigo usually comes from seeing how far you could fall. This kind comes from seeing how far you might be able to go. The feeling comes from your reachable world expanding faster than your mind can travel to its edges. It does not mean every direction is worth pursuing. It means you can no longer rely on the old map to tell you how far you can reach. Humans have always known the dizziness of possibility. What feels new is the boundary moving repeatedly, and from underneath you. Normally, the dizziness should fade as the map catches up — unless the map never gets the time. Capability changes. Intuition begins to stabilize. Then the world moves again. Perhaps possibility vertigo is not simply the temporary shock of acquiring a powerful new tool. It is what happens when your reachable world keeps expanding faster than you can redraw its map. The world gets larger. The old edges become harder to trust. And somewhere beyond them sits a question: What else have I called impossible simply because I could not see a path? — Van ## A System Nobody Chose September 6, 2026 · 7 min read · https://www.projectnothing.ai/essays/a-system-nobody-chose Every change can be correct. The system they produce can still be wrong. A thousand correct changes can build a system nobody chose. Engineers have always built upon things we do not understand. The compiler, the kernel, the database, the cloud, the orchestrator, the fifty libraries in the lockfile — nobody on your team understands them all the way down, and nothing about that is a crisis. It is how the profession works. So when someone says AI is letting us build things we don’t understand, the correct first response is: we always have. But the old bargains had a shape. Abstraction let us stop understanding what was below the boundary. AI lets us cross boundaries before we know which ones matter — or notice we crossed them. A compiler gives you a stable layer to stand on. An agent notices a production symptom, reads the application code, the Terraform, the IAM policy, the CI configuration and the logs, infers a relationship across them, changes two layers, and verifies the fix — moving through boundaries as though they were terrain. You no longer need to know which layer contained the problem, which abstraction applied — or, sometimes, what the problem even was. So the new thing is not that we can use systems we do not understand. We have always done that. The new thing is that we can now change them without ever learning what changing them used to require. Software has always outlived the understanding of its builders. Now it can be born without it. Before deciding how to feel about that, be honest about the upside. Suppose an AI builds a service. Nobody deeply understands its deployment machinery. It runs for ten years. Agents upgrade it, repair its incidents, tune its costs, patch its vulnerabilities. Ask the uncomfortable question: did anyone actually need to understand it? Maybe no. And that answer should not be waved away, because it suggests something engineers rarely say out loud: much of the knowledge we have traditionally carried was simply the cost of getting things done — knowledge we were forced to hold because the tools were not capable enough to hold it for us. If the tools can hold it now, putting it down is not a loss. For the generated glue code, the disposable dashboard, the migration tool that runs once — it may be pure gain. Then where is the problem? Understanding matters when “it works” stops answering the important question. A service returning 200s is easy. A test suite passing is easy. An agent confirming its change fixed the incident is easy. None of them answers: why does this service own this responsibility now? Why does this dependency exist? Why did the fix cross four architectural boundaries? Is this still the system we meant to build? Every individual change can pass every check it was given, and the checks can all be right. Every change can be correct. The system they produce can still be wrong. A thousand correct changes can build a system nobody chose. Understanding used to be the price of action: you could not make the change without paying it. While it was required, there was little reason to ask what else it was for. Now the machine pays the price for us — and for the first time, we get to ask. You don’t preserve comprehension for its own sake. You preserve the possibility of judgment. Here is the moment that matters. An agent reports its fix is correct: service A now publishes directly to queue C. The tests pass. The incident closes. And someone on the team says: service A was never supposed to know queue C exists. Nothing in the toolchain could have produced that sentence. Not more tests — they check what the system does, not what it should be. Not a second agent — it would confirm the first. The machine is correct about what the system does. The person disagrees with what the system has become. Every team carries knowledge like this: the boundary nobody crosses because crossing it once cost two teams a quarter of untangling; the dependency everyone treats as temporary; the module you may rewrite freely so long as it never touches the money path. Almost none of it is written down — until now, it never had to be, because the person who knew it was also the person making the change. When changes start arriving from a machine, that knowledge has to live somewhere outside the people who happen to hold it. Not as documentation of what was decided — as standing reasons to reject what works. Yes, it works. And we still shouldn’t ship it. Whoever can say why is exercising the judgment the machine did not replace. I am not describing this from above the transition. I am inside it. Some time ago I began accepting AI-proposed changes whose implementation I did not fully understand before accepting them — surface-tested, interrogated when they seemed consequential, but accepted. My bargain was simple: I will give up continuous understanding in exchange for dramatically greater capability, provided I keep a route back into understanding when it matters. So I built the route: machinery that wired together our production MCP servers, Slack, the documentation, the codebases, so that the reasoning behind any change could be reconstructed on demand. I wrote the how-to channel. I posted another company’s success story as encouragement. Then I released it, and the response was one offhand question — “What can you do?” — a concern that it was too expensive, a suggestion that everyone just use their own connectors, and then quiet. The non-developers I invited never tested it. Nobody asked. For a while I read this as a team not seeing what was coming. The likelier reading is less flattering to me: the tool solved the consequence of a bargain the organization had not agreed to make. Their implicit contract was still the old one — the person changing the thing should understand the thing — and under that contract, a machine for recovering understanding is an elaborate answer to a question nobody asked. And there is a worse reading still: perhaps AI had simply made it cheap enough for me to overengineer my own anxiety. Both can be true. I built infrastructure for a future that has not arrived, at a cost that would once have forced me to think harder about whether I should. This separation does not stop at tools. It is coming for seniority. AI doesn’t remove implementation from senior engineering. It removes difficult implementation as proof of senior engineering. Taking on the hardest problem used to bundle everything — depth, experience, architecture, failure intuition, judgment — into one visible act. Implementation difficulty made senior judgment visible, and AI is removing the lens. If an agent can produce the implementation, the difficulty stops hiding everything else we were calling seniority. For some, that will be uncomfortable. For others it will be overdue recognition: the parts of the work that never showed up in a diff — the cliff steered around before anyone saw it, the architecture carried in one head for years, the knowledge of what must remain true after the change — finally stand in plain view, with nothing in front of them. Ask yourself, without an audience: if producing the hard solution became easy tomorrow, what would still make you senior? Perhaps seniority becomes clearest in the moment when everything works and you are still the person who can say: we shouldn’t do this. The senior engineer still writes code. But increasingly, their harder job is making sure hundreds of individually sensible changes continue to describe one intentional system. There is a hole in all of this, and it is fair to name it rather than solve it. Today’s senior engineers know when understanding matters partly because they spent years being forced to understand. Tomorrow’s engineers may not get that training. We may be teaching people to skip the very experiences that taught us which things are safe to skip. That is not an argument for making juniors reproduce the old world out of educational purity. It is an admission that “load understanding only when it matters” contains a hidden problem: someone still has to learn what matters. What comes next is not a rule. It is a decision we now have to make deliberately. AI extends the distance we can travel before understanding becomes necessary. How far should we go? Sometimes all the way — some generated implementation may never need to be understood by a human, because understanding it contributes nothing. Sometimes we must stop and load: before a change crosses a domain boundary, when an agent invents a persistent dependency, when a subsystem starts acquiring responsibilities, when the architecture changes shape without anyone deciding it should. There is no universal threshold, and this essay will not pretend to supply one. The hard part is no longer understanding everything. It is knowing what you can safely leave unexplained. If understanding is no longer required to act, what is it for? Perhaps, increasingly, for the moments when success itself needs to be questioned. Understanding is what lets us disagree with success — what lets someone say: it works, and it’s wrong. Soon, changing a system may require almost no understanding of it. Engineering will increasingly be about knowing when that is perfectly fine — and recognizing the moment it isn’t. We are about to find out what engineering expertise means when you can change a system correctly without understanding it. Knowing when understanding still matters is about to become the discipline itself. — Van ## Superhuman and Redundant September 5, 2026 · 5 min read · https://www.projectnothing.ai/essays/superhuman-and-redundant What happens when the thing that made you feel useful can be summoned in seconds by someone who never had to become you? What happens when the thing that made you feel useful can be summoned in seconds by someone who never had to become you? Of course your experience matters. The years were real. They gave you instincts the newcomer does not have, context they cannot summon, mistakes they never had to survive. But that is less reassuring than it sounds. What if more and more of what twenty years of experience produces can now be had without the twenty years? “I am useful because I can do difficult things” — no one adopts that sentence in a single moment. It accumulates. You became the person people called when something was hard. Confusion became clarity in your hands. Problems that defeated others began yielding to you. Somewhere along the way, what you could do became the proof of who you were. The collapse reached me from its generous side first. I had been pulling at the phrase “AI slop,” and caught myself thinking: there is a book in this. And the horizon did not contract into everything I lacked — the years of training, the time, all the philosophy I had never read. It exploded. Of course I could write the book. And build a tool to help write it, and tools to research it, test its arguments, find every philosopher who had ever stood near its ideas. Things locked for decades behind insufficient capacity were suddenly all there, at once, waiting only to be asked for. And underneath the exhilaration, a quieter recognition: the horizon had not opened for me. It had opened for everyone. Both halves are true at once. You may become more capable than you have ever been, at exactly the moment capability counts for less and less. The machine can make you feel superhuman and redundant at the same time. And the cruelty is precise, because the work does not vanish. The code still needs writing. The diagnosis still needs making. The decision still needs reaching. The task survives. Your necessity does not. The obvious escape is to climb. Execution is cheap, so become the one who judges: develop taste, learn to select, be the person who knows what should exist. Ten thousand solutions can be generated; someone still has to say — that one, not those. But selection is not a sanctuary. The machines are coming for that too. We keep climbing the abstraction ladder, expecting somewhere above us a rung labeled HUMAN. Perhaps there is no such rung. None of this is unprecedented; technology has been making painstaking expertise cheap since the loom. What feels different is that this machine follows us when we move. This may still be an old story. But the interval between “become something else” and “that too” is shrinking, and somewhere in that shrinking, the oldest response to insecurity stops working. Maybe the future isn’t asking me to become better at what I do. Maybe it is removing the reason anyone needed those years from me. There will probably still be work, even when machines do nearly everything better. But having work is not the same as having the scarce ability that once made you difficult to replace. That scarcity was a moat: it protected your wage, your bargaining power, and the sense that the years you spent becoming good at something had bought you a place that couldn’t easily be taken away. We didn’t confuse productivity with worth for no reason. Work was not just proving worth; it was purchasing freedom — independence, status, the right to direct your own life. Which is why “your worth was never your work” is true and useless in the same breath. Take away the work and your worth survives. The freedom it was paying for does not — not unless something else picks up the bill. Start with the part you can still contest. If AI lets you do a week’s work in a day, do you get four days of life back — or five times as much work? The machine can return enormous capability to a person without returning control over where that capability is spent. Your employer can claim the saved time as more output. The market can claim it as a higher baseline. Your own ambition can claim it before anyone else does. When the machine saves you time, notice who claims the saved time. The thing you learned to do and the person you became while learning it are not the same thing. Maybe for you it was programming, or law, or design. Whatever the craft, the years also taught you to sit with confusion, take large problems apart, persist through failure, distrust elegant answers that don’t survive contact with reality. The market can cheapen what you produce. It cannot take back the judgment, patience, and instincts that producing it built in you. Those may not preserve your old place. They are what you carry into the next one. The same goes for the advice we pass down. We all inherited some version of it: become good enough at something that the world will need you. Decent advice in a scarcity economy. Psychologically dangerous in an abundance economy. But becoming capable can mean more than constructing a moat. Capability gives you more ways to encounter reality, more ways to participate in it, and more authority over what you bring into existence — even when it stops making you rare. For Monday, the advice is unromantic. Adapt aggressively. Use the machines. Stay close to consequential decisions. Protect your income. But do not surrender every gain in capacity to expectation, and do not make the next moat your identity. Don’t build a life around becoming impossible to replace. Build one where you still decide what deserves your attention and effort. And notice what the explosion felt like, before the fear arrived. Things locked behind insufficient capacity — ideas that could never get through the bottleneck of craft and time and confidence — can suddenly enter reality. For all that AI makes easier to replace, it also makes more of us expressible. The machine can generate, rank, and recommend possibilities. None of them becomes a life until you live it. If capability becomes abundant, the question is no longer how much more we can do. It is whether any of that abundance gives us greater authorship over our own lives. The world may stop needing some of what you became good at. That loss is real, and nothing here pretends otherwise. But you were never only what the world needed from you — and the person those years made now has more reach than at any point in your life. Go find out what you make when no one needs you to. — Van ## The New Minimum September 4, 2026 · 3 min read · https://www.projectnothing.ai/essays/the-new-minimum Once doing more becomes easy, not doing more begins to require an explanation. The frontier can move on its own. The minimum should not. “I haven’t really looked into it.” For most of my life, that sentence was explanation enough. Of course you hadn’t looked into it — looking into it cost an afternoon. Now a competent briefing is thirty seconds away, and I can feel the sentence changing in my mouth: less a fact about my time, more a confession about my choices. Nobody demanded this. No rule was announced. The cost of knowing collapsed, and what counts as reasonable moved with it. It isn’t only knowing. You buy something without comparing twenty alternatives. You hear a claim without checking it. You enter a meeting without asking for the background first. You send the three-line reply instead of the polished one a machine could write in ten seconds. None of these used to be failures, and most of them still aren’t. But they can begin to look like failures once the cost of doing more approaches zero. No one has to decide this. A capability becomes cheap. Enough people use it. Polished becomes ordinary; thorough becomes ordinary. Your unchanged work now sits beside work the machines helped polish, and the difference invites a story: they didn’t bother. They came unprepared. They could have checked. Nothing about your reply got worse. It simply fell below a standard that didn’t exist before. This is how the new minimum moves: once doing more becomes easy, not doing more begins to require an explanation. When extra effort was expensive, it usually had to justify itself. “I didn’t have time” protected enormous parts of life from optimization. As that excuse dissolves, omission starts to look less like limitation and more like choice. And choice is much easier to judge than limitation: “I couldn’t” earns sympathy; “I didn’t” invites judgment. But the machines do not decide what we owe each other. They expand what is possible. What is expected of us is set by people — by what we praise, what we excuse, what we hold against each other — and nothing about that is a law of nature. And “could have done more” is not the same as “more would have mattered”: every reply could be longer, every decision more researched, every document polished once more, and long before the machines run out of more, the improvements stop having a meaningful impact. What is possible is becoming bottomless. What we owe each other is not. Some minimums should rise. I no longer guess at things I can check in seconds — that excuse has genuinely expired. The mistake is not rising standards. It is letting the existence of a capability decide, all by itself, which standards rise — treating “easy now” as if it meant “required now.” So here is a small discipline for the next time AI makes something newly possible: don’t only ask, “Why wouldn’t I do this?” Ask, “What would justify making this the new minimum?” And when what you could have done starts to make you feel inadequate, hold the two sentences apart: Yes, I could have done more. No, that fact alone does not establish that I should have. The frontier can move on its own. The minimum should not. — Van # Experiments ## Assay Up · September 7, 2026 · https://www.projectnothing.ai/e/assay Tap the passages of an AI answer that matter, annotate them in a few words, and send them back as one deep-dive prompt. Nothing leaves the device. Tap a word in a ChatGPT or Claude reply, widen it to the sentence or paragraph, collect a few, annotate them, and send them back as one prompt — or save the whole conversation as a file. For anyone who reads AI answers slowly, on a phone, and is tired of copy-pasting the good bits back in. ## Horizon Up · September 5, 2026 · https://www.projectnothing.ai/e/horizon A longer horizon, a clearer view and a faster walk are all lying on the ground to be picked up. Attention is not. Walk an endless world, picking range, clarity and speed up off the ground. Go fast enough and you meet far more than you can reach. For anyone with more tabs, tools and options than they have hours. ## Verbatim Up · September 1, 2026 · https://www.projectnothing.ai/e/verbatim Paste a ChatGPT share link. Get the conversation back as a Markdown or plain text file. That is the whole product. Paste a ChatGPT share link. The conversation comes back as a Markdown or plain text file, both speakers in order. For anyone who wanted to keep a conversation and found they only had a URL. ## Needed Up · August 31, 2026 · https://www.projectnothing.ai/e/needed A stranger left a question. A machine has already answered it — correctly, fully, for free. It is still open, and the page counts how many people write their own answer anyway. Answer a stranger’s question in a few sentences, next to the answer a machine has already given. For anyone who still likes being useful to a person. ## Slop Up · August 30, 2026 · https://www.projectnothing.ai/e/slop Judge five things without being told where they came from. Then be told. The page keeps score of how often the provenance changed your mind. Judge five pieces of writing, art and code before you are told where each came from. Then see what changed. For anyone who has called something slop without checking. ## Done Up · August 30, 2026 · https://www.projectnothing.ai/e/done Tell it the thing that has been mildly annoying you for a month. One costed route, one part, a date — then it comes back to ask whether it worked. Describe one household problem you have been ignoring. Get one costed route, one part and one date, then a check-in. For people in Stockholm with a month-old nag on the list. ## Up Peak Up · August 30, 2026 · https://www.projectnothing.ai/e/up-peak Four lifts, twenty floors, one morning rush. Change one thing and the identical morning runs again. Watch four lifts serve a morning rush. Change one thing about the building and run the same morning again. For anyone who has stood in a lobby wondering why the lift takes so long. ## Crawled Up · August 30, 2026 · https://www.projectnothing.ai/e/crawled Every machine that reads this website, what it takes, and how long after publishing it turns up. A live ledger of which machines read this site, what they took, and how long after publishing they arrived. For anyone publishing on the web and wondering who, or what, is reading. ## Out of Ten Up · August 30, 2026 · https://www.projectnothing.ai/e/out-of-ten Give a stranger a mark out of ten in one second, then keep looking until the mark cannot hold them. Give a stranger a mark out of ten in one second. Then keep looking, until the mark stops fitting. For anyone honest enough to admit they rate people on sight. ## Perfect Score Up · August 29, 2026 · https://www.projectnothing.ai/e/perfect-score Forty-two of the greatest games ever made, taken apart into ten ingredients you then have to buy. Forty-two of the greatest games ever made, taken apart into ten ingredients. You have 500 points to rebuild one. For anyone who has argued about the greatest game of all time. ## Insula Up · August 28, 2026 · https://www.projectnothing.ai/e/insula A running argument about what the old city builders got right, made out of cities that simulate. Small cities that actually run, each making one argument about what the 1998 city builders got right. For anyone who still thinks about Caesar III. ## Exactly Up · August 26, 2026 · https://www.projectnothing.ai/e/exactly Say what a song did to you, and read what a stranger wrote about the same one. Write one sentence about what a song did to you. Read what a stranger wrote about the same song. For anyone who has ever needed to say it to somebody. ## Spent Up · August 25, 2026 · https://www.projectnothing.ai/e/spent Your calendar, read the way a bank reads your spending. Drop in a month of your calendar. Read it back as a statement of where the hours actually went. For anyone whose week is booked solid and cannot say by what. ## Sieve Up · August 24, 2026 · https://www.projectnothing.ai/e/sieve Type a website address and find out where it is losing customers. Type in a website address. A free scan comes back with where it is losing customers. For anyone who runs a site and suspects the leak is somewhere obvious. ## Hold the Note Up · August 22, 2026 · https://www.projectnothing.ai/e/hold-the-note A melody that refuses to continue until you sing the note. A melody that stops and waits until you sing the note in tune. Then it continues. For anyone who sings in the car and wonders if they are in tune. ## Built From a Phone Up · May 4, 2026 · https://www.projectnothing.ai/e/built-from-a-phone A terminal, an agent, a deploy pipeline, and no laptop. The story of a 2D Godot game built and shipped from a phone terminal, without opening a laptop. For anyone building in the pockets of time between everything else. ## Instruction Elasticity Index Up · May 4, 2026 · https://www.projectnothing.ai/e/instruction-elasticity How far five frontier models bend before they break, measured every week. Five frontier models run the same tasks at four levels of prompt scaffolding, every week: how far each bends before it breaks. For anyone writing prompts and wondering how much of it matters. ## Subscribe to Nothing Up · May 4, 2026 · https://www.projectnothing.ai/e/subscribe-to-nothing A paid subscription that delivers, and has always delivered, nothing. A monthly subscription that delivers nothing, on purpose. The money funds the experiments on this page. For anyone who would rather fund the process than buy a perk. ## On Me Being built · August 30, 2026 · https://www.projectnothing.ai/e/on-me One photo in, and eight ordinary garments come back — four on a model, four on you, shuffled and never side by side. The page counts which ones you said you would wear. One photo of you. Eight ordinary garments come back, four on a model and four on you, and the page counts which you would wear. For anyone who has never once looked like the model. ## Make Interesting Became a product · May 4, 2026 · https://www.projectnothing.ai/e/make-interesting Daily social drafts in your voice, from the signal you already read. Drop in a URL. Every morning, social posts drafted in your voice from the day’s discourse, ready to approve. For founders who should post and never find the hour. Now living as Make Interesting. ## Signal Weather Gone · May 20, 2026 · https://www.projectnothing.ai/e/signal-weather A daily map of what is loud, what is early, and what is about to collide. A daily map of the discourse: what is loud, what is early, what is about to collide. Runs inside the studio. For the people running this place. Not enough engagement. A daily chart of the discourse that nobody measurably looked at; retired with the rest of the scheduled social output. ## AI-deas Gone · May 11, 2026 · https://www.projectnothing.ai/e/ai-deas One buildable product idea a day, generated from the raw signal pool. One buildable product idea a day, generated from the studio’s signal pool. It feeds the idea queue here. For the people running this place. Not enough engagement. The ideas fed a queue nobody was drawing from, and the daily posts that carried them are retired with the rest of the scheduled output. ## Mutation Observer Gone · May 4, 2026 · https://www.projectnothing.ai/e/mutation-observer A landing page that rewrites itself and grades its own conversions. A landing page that rewrote its own copy and graded its own conversions. For the people running this place. Never left the idea. It sat on the experiments page marked "coming soon" for three months, which is its own kind of result: the traffic was never going to be enough to attribute anything. ## Micro-Break Gone · May 4, 2026 · https://www.projectnothing.ai/e/micro-break The gap between what people wanted and what they got, in two lines. Two-line posts on the gap between what people wanted and what they got. A distribution test run from inside the studio. For the people running this place. Not enough engagement. Two-line hooks went out every day for four months without a measured reader, and the pipeline behind them is on ice with the portal. ## Code of Laws Gone · May 4, 2026 · https://www.projectnothing.ai/e/code-of-laws A day of AI discourse, compressed into one law. Every day at 14:00 UTC, the day’s AI discourse from Hacker News, Reddit and arXiv, compressed into one law. For anyone who wants the day’s argument in one sentence. Not enough engagement. One law a day went out to every connected network for four months without a measured reader — and the studio has stopped scheduled social output altogether. ## LLM Price Volatility Gone · May 4, 2026 · https://www.projectnothing.ai/e/llm-price-volatility What a million tokens cost, tracked until the line stops falling. The price of a million tokens across frontier models, charted since 2021 on a log scale and updated weekly. For anyone budgeting for AI, or betting on where the price goes. Not enough engagement. The chart was cheap to keep and nobody visibly cited it; the studio would rather write the essay than maintain the reference. ## Voidle Gone · May 4, 2026 · https://www.projectnothing.ai/e/voidle A puzzle game where growth is the failure mode. Achieve nothing. A sliding-tile puzzle played backwards: merge tiles past a threshold and they vanish. Empty the board to win. For anyone with five minutes and a 2048 habit. Not enough engagement. Four months up, unmeasured, and nobody visibly came back to play it. ## Velocity Cards Gone · May 4, 2026 · https://www.projectnothing.ai/e/velocity-cards Repositories heating up, rendered as something you would actually post. Repositories gaining speed on GitHub, rendered daily as cards you could post. Acceleration, not popularity. For anyone who wants to see what is about to be popular, not what already is. Not enough engagement. Its daily card existed to feed the Signal from the Void post, and it drew no measurable attention of its own; it goes with the cron that produced it. ## The Replacement Index Gone · May 4, 2026 · https://www.projectnothing.ai/e/the-replacement-index Twelve questions, four axes, sixteen ways to be worried about your job. Twelve scenario questions map you to one of sixteen developer archetypes, with the blind spots that come with each. For developers wondering, honestly, how replaceable they are. Not enough engagement. Four months on the site without a measured funnel and without a share that anyone traced back to it. ## Signal from the Void Gone · May 4, 2026 · https://www.projectnothing.ai/e/signal-from-the-void One developer insight a day, extracted from whatever GitHub is quietly obsessed with. Every day at 05:00 UTC an agent reads GitHub’s trending page and publishes the one thing worth saying about it. For anyone who wants the trending page read for them. Not enough engagement. It posted a daily insight to every connected network for four months, unmeasured, to an audience that never visibly grew — and the studio has stopped scheduled social output altogether. Last generated: 2026-09-26T09:21:11.072Z