# I Got 99 Essays, But a Throughline Ain't One *I let the machines read everything I've written since 2023. They found five things I keep saying.* By [Nye's Digital Lab](https://paragraph.com/@nyewarburton.eth) · 2026-10-04 essays, analysis, recap, writing --- > _Nye's Digital Lab is a weekly scribble on creativity in an age of rapid change._ _This week I'm reading myself. I have carefully avoided doing that for three years and 129,000 words._ * * * [Subscribe](https://paragraph.com/@nyewarburton.eth/subscribe) ![](https://storage.googleapis.com/papyrus_images/cd6aac89de84451aa7c028d8073b428cede8da5e269e315d8e46566905d7b3e3.png) _High level analysis of my library of essays. Source Claude Opus 5.5._ **Essay #100!** ### The Robots Read Me This week is my 100th essay, so... I let the robots read me. Not one essay. **All of them.** About 129,000 words. I split the pile between four models, gave each one twenty-five or so essays, and asked one question: * * * > _What does this guy keep saying?_ * * * I wanted a pattern. Some sign of what **Nye's Digital Lab** is actually about. I hate rereading my essays. I write them, I publish them on Sunday, **and I move on**. Going back feels like hearing my voice on a voicemail. I did read some of them. But, I had the machines do most of it. And the first thing they showed me wasn't an idea. It was a calendar. **The Digital Lab** didn't start as a weekly practice. It started as a blog in January 2023. I wrote eight essays that year, and then went quiet for eight months. In 2024 I wrote seven, with three gaps longer than two and a half months. Those first fifteen essays took me almost two years. Then, on March 30, 2025, something changed. I published _Here Comes Reinforcement Learning_, and I haven't gone more than two weeks without an essay since. **Every Sunday of 2026 has one.** The blog years were me waiting until I had something worth saying. The weekly years are me saying something, and finding out on the page if it was worth it. I am literally just jumping ahead of myself and **pushing publish.** Anyway, here are the five ideas I seem to talk about the most in 99 essays of **Nye's Digital Lab.** * * * The Five Ideas ============== * * * **One. Judgment survives.** --------------------------- _When execution is free, judgment is what's left._ (19 essays) It's the DJ who doesn't just drop the needle. It's "_you do not create with AI, you search._ " It's the hours migrating from the fingers to the ear. Last week, I said it was whoever people trust to pick the shot. I've said this in every lane I write in, in three different vocabularies. It's my most repeated idea, and maybe my most self-serving. According to AI, I almost never say _taste_. I say **trust**. Whoever people _trust_ to pick the shot. Maybe judgment isn't the skill. Maybe it's the reputation you build by showing up. The best example: * [**Helmets and Violins**](https://paragraph.com/@nyewarburton.eth/helmets-and-violins) — **Judgment.** The frontier models are the biggest sampler ever built, and the hours migrate from the fingers to the ear. * * * **Two. You have to earn it.** ----------------------------- _The friction was the education._ (15 essays) Count before the calculator. Beat the boss without the automation. Sand the friction away and you get a faster tourist. I fully wrote one of these by hand, in August, _in a cold sweat._ I think this is the whole academic crisis in one line. We don't need more AI literacy. * * * > We need to teach people **the grit to survive what's coming.** * * * * [**The Graphics Hack**](https://paragraph.com/@nyewarburton.eth/the-graphics-hack) — **Earn it.** Two doors, and I'm standing between them. * * * **Three. Open beats closed.** ----------------------------- _The edge beats the cloud, and the bubble pops._ (23 essays) My widest thread, and really two bets wearing one coat. The first half held. Open models sit a few months behind the frontier. A fixed unit of intelligence gets ten times cheaper every year. Six in ten tokens on the biggest marketplace now run on Chinese open weights. The second half, the collapse, I've been calling since _Universal Basic Compute_. The capex I called a bubble in April doubled by September. And we all seem to still run Claude Code every single day. I still think the big bubble goes _boom_. I just don't know when. * * * * [**The Big Bang**](https://paragraph.com/@nyewarburton.eth/the-big-bang) — **Open beats closed.** The bubble I wanted to pop, and it hasn't. * * * **Four. The work moves.** ------------------------- _It doesn't disappear, and the schools that stall get merged._ (18 essays) Detroit, then Hollywood. The sawed-off first rung. Merge before you're merged. The _gone_ half of this claim is the best-evidenced thing I've written. The _moved_ half I've asserted eighteen times and shown zero. Bangalore. Micro-studios. Utility players. Those are the places I said the work would go. The one number I actually checked cuts the other way. Twenty thousand games on Steam last year. Six hundred got a thousand reviews. * * * * [**My Heart Bleeds for Hollywood**](https://paragraph.com/@nyewarburton.eth/my-heart-bleeds-for-hollywood) — **The work moves.** Detroit, one industry earlier. * * * **Five. The ledger pays artists.** ---------------------------------- _Attribution by name will pay the artists, and the ledger is how._ (15 essays) This was actually the reason I started writing. I felt that artists should be compensated for their ideas. Six of my first eight essays. I've apologized for it in print at least seven times, and re-prescribed it every single time. Yes, it's true. I believe in the blockchain. Aaaaand ... I feel really bad about it. Notice this essay is cryptographically secured on Paragraph, a blockchain based platform. I am here for a reason. * * * * [**The Ghost of Napster**](https://paragraph.com/@nyewarburton.eth/the-ghost-of-napster) — **The ledger.** The founding claim, with its dated predictions attached. * * * [Share](https://paragraph.com/@nyewarburton.eth/rf6AF741C6ONKDXPgqf2) * * * And The Analysis ---------------- Here's some of the data. The machines also built me a map! Every dot is an essay. Essays that use similar words sit close together. ![](https://storage.googleapis.com/papyrus_images/56f962d6181a27b8d965068cae5ca80828c8969bba43659ed400c5bb452a8c19.png) The top panel colors each essay by the folder I filed it in. The bottom colors it by what the text actually says. Where they disagree, my filing and my writing disagree. > Here's the coolest stuff (I think). **There are no walls.** The essays don't split into clean topics. **Teaching and personal are the same lane.** When you read the words, my education essays and my personal essays blend together. The teaching essays _are_ the personal essays. I didn't know that! **"AI" is not a topic.** My AI & Technology essays don't resemble each other any more than they resemble everything else. AI is a constant. **Only robots stick together.** Robotics is the one folder where the essays actually sound alike. Everything else wanders. **Almost a third of the essays live in the wrong house.** For 30 of 97 essays, the three closest essays by text are all filed somewhere else. Personal Reflection is the worst. **I write in runs.** Consecutive essays are more alike than random pairs. The Sunday essay is usually a continuation of last Sunday's. So, apparently _I am literally all over the map._ * * * Just Keep Showing Up -------------------- So I have 99 essays, and a throughline ain't one. I used three major AI models, a similarity map, and a pile of tables to find out what I believe. **The answer was in the calendar.** Nothing got clear when I had a big idea. It got clear when I started writing every week. You need to do the work. That's it. That's the throughline. _Show up on Sunday. Write the thing. Do it again._ I was wrong about Jay-Z. Maybe it's the other guy: Woody Allen. 80% of success is showing up. I'll settle for every week if I can keep this up. Until next time. **_Make it Happen._** * * * Colophon -------- This essay was constructed from a deep dive analysis of the past 99 essays from **Nye's Digital Lab** using Claude Fable 5.1, Gemini 3 Pro / 3.1 Pro, and Deep Seek v4. Additional revisions were done by a human pushing buttons on a keyboard. Nye Warburton is an artist and educator from Savannah, GA. For more information: [nyewarburton.com](http://nyewarburton.com) * * * [Subscribe](https://paragraph.com/@nyewarburton.eth/subscribe) --- *Originally published on [Nye's Digital Lab](https://paragraph.com/@nyewarburton.eth/i-got-99-essays-but-a-throughline-aint-one)*