Former YC CTO. Startup CEO. Engineering leader. AI researcher studying how humans compress reality into stories. Published author. Substack ↓ he/they 🧡
I work in AI every day. I build with it, research it, write about it. I also come from three generations of educators and I'm raising a ten-year-old.
The evidence says we're about to make a catastrophic mistake putting AI in classrooms. A 17-point comprehension gap. A −0.68 correlation between AI use and critical thinking. The youngest users show the highest dependence.
Less screen. More human. Better outcomes. The data is not complicated.
https://t.co/vcBSm5VMob 🧡
Day 4 of my narrative compression series: the convergent evolution argument.
12 unrelated domains — case law, clinical medicine, military AARs, open-source software, religious traditions — all independently evolved narrative as their knowledge format.
Not coincidence. Selection pressure. When you need to transmit understanding under bandwidth constraints, narrative is the only format that carries causal structure.
🧡 https://t.co/bN60NdrFqC
400 developers. Same AI tool. Opposite reactions.
The 60-year-olds can't sleep from excitement. The principal engineers feel erased.
Both are right. The distinction nobody is naming: prosthetic tools replace capacity. Therapeutic tools restore it. Same tool can be either depending on who's holding it.
Wrote about it: https://t.co/kzvY3qe5bi
New essay: why we're information-rich and understanding-poor — and why current AI is the problem's most sophisticated expression.
Every major platform is optimizing for recall. Longer context. Better retrieval. More memory. But the bottleneck was never recall. The bottleneck is integration.
Summarization is just shorter accumulation. It reduces volume without changing format. You need something that changes the ratio of structure to noise entirely.
There's a term for that ratio. It's called inferential yield. That's tomorrow.
Day 1 → https://t.co/BsY2SrLioG
🧡
I just published something I've been building toward for two years.
7 Days of Understanding Narrative Compression — a daily series walking through why we're drowning in information but starving for understanding, and what narrative has to do with the fix.
Two research papers. Twelve domains. One week.
Day 1 drops tomorrow. Subscribe if you want it in your inbox.
https://t.co/ouUNcdWPrL
🧡
Six months sober today. 181 days. I wrote the book before I’d been sober this long — last time I made it a year and a half. Don’t miss it a bit. Feels good to catch up. 🧡
RedDep isn't just consulting. It's a product thesis.
AI doesn't need more memory. It needs sleep. The computational equivalent — select, compress, integrate, dream.
I call it emplotment-as-service. No one else is building it because no one else is using narrative compression as a formal framework.
The consulting is where research meets reality. But it's the starting surface, not the destination.
Full essay → https://t.co/bfciVtUkQn
https://t.co/dhkNHqd4zA 🧡
What happens when you book a session with RedDep? I don't hand you a document. I leave you understanding something you couldn't see alone.
https://t.co/XaK5cwHVnM 🧡
Tres niveles. Un principio. Cada sesión de RedDep termina con más capacidad, no más dependencia.
Asesoría. Inmersión técnica. Estrategia ejecutiva. La pregunta siempre es la misma: ¿sales más fuerte?
La sesión termina. La capacidad se queda. 🧡
Lee el ensayo completo → https://t.co/nUytxohvut
https://t.co/dhkNHqd4zA
Your AI remembers everything and learns nothing.
Your brain does the opposite — it prunes every night, selectively forgetting what stopped mattering so the signal gets sharper.
No AI system does this. They just hoard. Bigger windows, more RAG, total retention. No integration.
1/2
Your AI remembers everything. That's not a feature — it's the bug.
Selective forgetting is what converts experience into understanding. Your brain prunes every night, weakening what stopped being relevant so the signal gets stronger.
AI does the opposite. Bigger context windows. More RAG. Total retention. No pruning.
That's why your assistant gives technically correct answers that feel hollow. It's not missing intelligence — it's missing sleep.
New essay on what narrative compression could fix 👇
https://t.co/guhDHAoLUO 🧡
De Noche at La Fondita is easily one of the best meals you can have in PDX. From the start of silky smooth Caldo de Vegetales, we knew we were in for a treat. The Quesadilla Oaxaqueña with huitaloche and the Empanadas with chanterelle mushrooms were absolutely spectacular. Such a treat! Definitely going back and highly recommend.
Anthony Bourdain’s Lasagne Bolognese. Hours of work, but totally worth it. The depth of flavor and perfect textures of soft noodles and good crunchy and chewy bits made for an incredible bite of food. Thanks Anthony. Miss you. #lasagna#anthonybourdain#homechef