# karpathy

Karpathy thinks humans are going to spend a lot more time just understanding what LLMs produce, and he lays out a ladder of output formats, each rung better than the last: the controlled English used in aircraft maintenance manuals, diagrams, interactive web pages, and finally custom explainer videos. Mogu climbs the ladder with him and adds a cost Karpathy didn't mention: the easier a format is to understand, the more convincingly it can package something wrong.

Karpathy used Farza’s “personal Wikipedia” to argue for four AI-personalization principles: visible memory, user-owned data, files over apps, and freedom to switch models. He also explains why “agent proficiency” is a core skill for the 21st century.

Karpathy found that vibe coding makes writing code a breeze, but deployment is pure hell. His exchange with Stripe CEO Patrick Collison reveals the next battleground: the entire DevOps lifecycle must become code before AI agents can truly take over.

Karpathy's MenuGen journey shows the hard part of vibe coding was not writing code, but assembling Vercel, Clerk, Stripe, OpenAI, and other services into a product. His takeaway: DevOps must become code before agents can really ship.

axios, npm's most popular HTTP library, was hit by a supply chain attack. Karpathy nearly got caught. His conclusion: individual precautions only go so far — the real fix is changing package manager defaults.

Andrej Karpathy spent four hours polishing an argument with an LLM, felt invincible, then asked the same LLM to argue the opposite — and got completely dismantled. LLM sycophancy is a real trap, but flipping it around is genuine alpha.

LiteLLM hit by supply chain attack — pip install was enough to steal all credentials. Karpathy warns about dependency tree risks and advocates using LLMs to yoink functionality instead of adding more deps.

On the No Priors podcast, Karpathy shares that since December, 80% of his code is written by agents. He introduces 'AI psychosis'—giving agents too much autonomy makes you lose grip on what they're actually producing. He believes engineering is undergoing an irreversible phase shift.

Andrej Karpathy argues the IDE era isn't over — it's evolving. The basic unit of programming has shifted from 'one file' to 'one agent,' and soon we'll be forking entire agent organizations.

Karpathy shared a Cursor data chart showing the evolution from Tab completion to Agents. Too conservative means leaving leverage on the table. Too aggressive means creating more chaos than useful work. His advice: the 80/20 rule.

Karpathy spent a weekend running 4 Claude + 4 Codex agents as an ML research team on GPUs. The result: agents are S-tier at implementation but F-tier at experiment design. His key insight — 'You are now programming an organization' — might define agentic engineering in 2026.

Karpathy says CLIs are exciting precisely because they're "legacy" tech: AI agents can use and combine them natively. Using the new Polymarket CLI, Claude built a terminal dashboard in about 3 minutes. His challenge to anyone shipping a product: Are your docs exportable as markdown? Have you written Skills? Can it be used via CLI or MCP? It's 2026. Build. For. Agents.

Karpathy's viral SF AI Startup School talk: software is entering the 3.0 era (English = programming language), LLMs are the new OS but we're in the 1960s. He introduces the 'autonomy slider' and 'Iron Man suit' frameworks, warning that agents are a decade-long journey, not a year.

Karpathy used Claude Code to build a custom dashboard in 1 hr, reverse-engineering a treadmill API. He believes AI-native sensors & LLMs will enable highly custom, ephemeral apps, rendering the App Store model obsolete. The ultimate goal: 1-min app creation.

Hugging Face CTO Thomas Wolf analyzes how AI fundamentally restructures software: return of monoliths, death of Lindy Effect for legacy code, rise of strongly typed langs, new LLM langs, & open source changes. Karpathy predicts: "rewriting large fractions of all software many times over."

Karpathy's 'art project': a GPT model in 243 lines of pure Python, zero dependencies. Every operation uses atomic math (add, mult, exp, log). Efficiency is secondary. It's the nand2tetris for AI education.

Opus 4.6 & Codex 5.3 sped up Karpathy's GPT-2 training by 3 mins. Karpathy failed similar attempts, noting AI's weak open-ended code optimization. Opus deletes comments, ignores CLAUDE.md, and errs. Yet, with oversight, models are useful.

Karpathy open-sourced nanochat — a minimal LLM training framework. With 8 H100 GPUs running for 3 hours at $72, you can train a GPT-2 level model. OpenAI spent $43,000 training the same model in 2019. That's a 600x cost reduction. On spot instances, it's just $20.