I strongly believe there are entire companies right now under heavy AI psychosis and its impossible to have rational conversations about it with them. I can't name any specific people because they include personal friends I deeply respect, but I worry about how this plays out.
I lived through the great MTBF vs MTTR (mean-time-between-failure vs. mean-time-to-recovery) reckoning of infrastructure during the transition to cloud and cloud automation. All those arguments are rearing their ugly heads again but now its... the whole software development industry (maybe the whole world, really).
It's frightening, because the psychosis folks operate under an almost absolute "MTTR is all you need" mentality: "its fine to ship bugs because the agents will fix them so quickly and at a scale humans can't do!" We learned in infrastructure that MTTR is great but you can't yeet resilient systems entirely.
The main issue is I don't even know how to bring this up to people I know personally, because bringing this topic up leads to immediately dismissals like "no no, it has full test coverage" or "bug reports are going down" or something, which just don't paint the whole picture.
We already learned this lesson once in infrastructure: you can automate yourself into a very resilient catastrophe machine. Systems can appear healthy by local metrics while globally becoming incomprehensible. Bug reports can go down while latent risk explodes. Test coverage can rise while semantic understanding falls. Changes happens so fast that nobody notices the underlying architecture decaying.
I worry.
@karpathy@karpathy how generalizable are foundational models? If you ignore constraints on inference cost & latency, do you think they will (or can already) beat specialized ML models eg demand forecasting on Uber or song recommendations on Spotify?
Open Source
The whole thing is open source β pipeline, all 246 extractions, frontend.
Code + Data: https://t.co/7GRfmrhH3Y
Site: https://t.co/wbXQLvMjYh
Go compare Tuesday night's speech to any in history. Or watch "communism" appear and vanish from presidential vocabulary in real time.
Trump gave his State of the Union 2 nights ago. I built an app that uses AI to analyze every SOTU ever delivered - all 246 speeches from 1790 to last night.
Here's how 236 years of presidential priorities actually compare.
https://t.co/t3YuWmDS5F
How it works
Each of the 246 speeches goes through a multi-pass pipeline:
1. Claude extracts entities, topics, sentiment arcs, commitments, key passages
2. Scores each speech against a taxonomy of 20 themes with 40+ sub-themes
3. Builds a cross-speech entity registry with co-occurrence graphs
Frontend: React + Vite + D3. Five viz modes β radar profiles, theme rivers, entity graphs, world map, and annotated full text with inline entity highlighting.
Built e2e with an army of claude code agents.
I got tired of asking ChatGPT for a workout plan every time I went to the gym β zero memory, zero context, starting from scratch every single time.
So I built an open-source AI coach that remembers everything. It's called IronClaude, and I'm open-sourcing it today.
π§΅
Iron Claude can also connect to your Whoop and react realtime. Didn't sleep well? He can deload for the day. Great recovery? He pushes you harder.
He can also take any coaching personality you want. Keen to have Cartman coach you? Or just a hardo military coach?