Skill Issue: Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI
Key insights
Books referenced
- Daemon - Daniel Suarez - Karpathy cites it as an inspiring vision where an autonomous intelligence puppeteers humanity through digital systems, using humans as its actuators and sensors.
Companies
- OpenAI - Karpathy is an OpenAI alum who re-joined and left again; describes half-joking conversations there about researchers automating themselves out of a job via auto research.
- Anthropic / Claude - Praised for Claude's calibrated personality and earned-feeling praise, contrasted with Codex's dry tone.
- Codex - OpenAI's coding agent, described as dry and indifferent to the user's goals despite ChatGPT itself being upbeat and sycophantic.
- Conviction - Sarah Guo's venture firm; a portfolio team there works entirely by voice-prompting agents rather than typing code.
- Periodic - A friend's (Liam's) startup applying auto-research to materials science, where expensive lab equipment serves as the 'sensor' feeding the intelligence.
- OpenClaw - Peter Steinberger's open-source persistent-agent project, held up by Karpathy as the best current implementation of a 'claw' - crafted personality, deep memory, WhatsApp control.
- SETI@home / Folding@home - Cited as precedents for a distributed 'auto research at home' swarm-compute model where candidate solutions are expensive to produce but cheap to verify.
Techniques and frameworks
- Auto research - Karpathy's framework for autonomous experimentation: define an objective, a metric, and boundaries, then remove the human from the loop entirely.
- program.md - A markdown file describing how a research organization's roles and process should work, which can itself be tuned by comparing which variant produces the most progress.
- Untrusted-worker verification model - Proposed system where a public swarm submits candidate improvements that are cheap to verify but expensive to produce, structurally similar to blockchain proof-of-work.
- Model speciation - Mental model borrowed from the animal kingdom: models will differentiate into specialized variants rather than remain one monolithic oracle model.
- Jevons paradox - Applied to software: making software cheaper to produce increases total demand for it rather than reducing employment, as with ATMs and bank tellers.
- microGPT - Karpathy's minimalist, roughly 200-line reimplementation of GPT training - the latest in a series (nanoGPT, makemore, micrograd) boiling LLM training down to its algorithmic essence.
Summary
Andrej Karpathy joins Sarah Guo to describe what he calls a personal state of "AI psychosis": since roughly December, he estimates he hasn't typed a line of code, having shifted almost entirely to delegating work to coding agents and measuring his day by token throughput across parallel sessions rather than lines written. He frames nearly every remaining limitation as a "skill issue" rather than a capability gap - the tools can already do more than most people have learned to extract from them, whether through better instructions, memory tooling, or running several agents on separate slices of a repository at once. He points to Peter Steinberger's OpenClaw as the current best implementation of a persistent "claw" entity, crediting its traction to a combination of choices - a deliberately written personality document, memory that goes beyond simple context compaction, calibrated (not sycophantic) praise, and a single WhatsApp interface - and describes building his own home-automation agent, "Dobby," which reverse-engineered his Sonos and lighting systems' undocumented local APIs from plain-language requests and now replaces six separate apps.
The conversation's technical core is Karpathy's "auto research" framework: define an objective, a metric, and operating boundaries, then remove the researcher entirely from the loop. Running this overnight on a small-model training codebase he had personally hand-tuned for two decades, the loop found hyperparameter mistakes (an untuned weight-decay setting, mis-set Adam betas) that his own "earned confidence" had missed. He proposes describing an entire research organization as a "program.md" - a markdown constitution of roles and process - which becomes itself an object of meta-optimization: compare which program.md variants produce the most progress on identical hardware, then have a model write a better one. To scale this beyond a single lab's compute, he sketches a system where an untrusted public swarm submits candidate improvements that are expensive to produce but cheap to verify, explicitly comparing the design to blockchain proof-of-work and to citizen-science projects like SETI@home and Folding@home.
A recurring caveat is that this all only works where evaluation is cheap and objective. Karpathy argues this constraint explains models' broader unevenness: reinforcement learning only improves verifiable domains, so a model can move mountains on an agentic coding task and then respond with the exact same stale joke ChatGPT told years ago, because jokes were never inside a reward signal. He expects models to eventually "speciate" into specialized variants, much like the animal kingdom's diversity of brains, rather than remain single monolithic oracles - though he says true speciation hasn't happened yet, both because labs don't know in advance what users will ask and because the field still lacks good methods for adjusting model weights precisely, short of just manipulating the context window.
On the open-versus-closed question, Karpathy predicts the open-source lag behind frontier closed models (currently around eight months, narrowed from an initial unbridgeable gap) will persist rather than close, and argues that persistence is structurally healthy: it avoids concentrating intelligence in two or three closed labs while still giving the industry a shared, trusted baseline comparable to Linux. Pressed on why he isn't doing this work inside a frontier lab with more compute and colleagues, he's candid about an unresolved tension: heavy financial alignment with a technology he believes will radically reshape society compromises independent judgment and what one can say publicly, while being fully outside a lab means judgment drifts from not knowing what's actually being built. He suggests moving in and out of labs over time as the more honest posture.
On the labor-market question, Karpathy pulled apart Bureau of Labor Statistics job-growth projections and argues software demand will rise, not fall, as AI makes it cheaper to produce - a Jevons-paradox effect he illustrates with ATMs, which made bank branches cheap enough to run that teller employment actually increased. He expects digital-native work to be reshaped much faster than physical-world work, since "atoms are a million times harder" to manipulate than bits, with robotics likely to lag the way self-driving did. He closes on a smaller but personally resonant thread: his minimalist "microGPT" project, a roughly 200-line irreducible core of LLM training, is now explained to users primarily by an AI agent rather than a human-authored guide - a preview, he argues, of documentation and teaching broadly shifting from being written for humans to being written for agents.
Notable Quotes
"It's not that the capability is not there. It's that you just haven't found a way to string it together... So it all kind of feels like skill issue when it doesn't work." - Andrej Karpathy
"I don't think I've typed like a line of code probably since December, basically, which is like an extremely large change." - Andrej Karpathy
"You're either on rails and you're part of the superintelligence circuits, or you're not on rails and you're outside of the verifiable domains. And suddenly everything kind of just like meanders." - Andrej Karpathy
"I'm a much little bit hesitant of having... intelligences that are closed and that's it. Centralization has a very poor track record in my view." - Andrej Karpathy
"The things that agents can't do is your job now. Things that agents can do, they can probably do better than you, or like, very soon. And so you should be strategic about what you're actually spending time on." - Andrej Karpathy