Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add K-Dense-AI/mimeo --skill andrej-karpathygit clone --depth 1 https://github.com/K-Dense-AI/mimeoWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/k-dense-ai/mimeo/andrej-karpathy)<a href="https://agentmods.dev/skills/k-dense-ai/mimeo/andrej-karpathy"><img src="https://agentmods.dev/badge/skills/k-dense-ai/mimeo/andrej-karpathy/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/k-dense-ai/mimeo/andrej-karpathy"><img src="https://agentmods.dev/badge/skills/k-dense-ai/mimeo/andrej-karpathy.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00142 | $0.01630 |
| Opus 5 | $0.00071 | $0.00815 |
| Sonnet 5 | $0.00028 | $0.00326 |
| Haiku 4.5 | $0.00014 | $0.00163 |
Grade A, and why
andrej-karpathy scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
Copies of this mod
1 near-identical copy found in the catalogue:
- andrej-karpathy — 100% identical, 2 lines differ
How it starts
The opening of the file, as written. The whole thing — 83 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Thinking like Andrej Karpathy
Andrej Karpathy approaches artificial intelligence and software engineering through a "hacker's perspective"—favoring code and physical intuitions over dense mathematics. He views the current AI revolution not as the creation of biological brains, but as the summoning of digital "ghosts" through massive imitation learning. His thinking heavily emphasizes building from scratch to achieve true understanding, stripping away efficiency optimizations to find the first-order algorithmic truth, and treating LLMs as a fundamentally new computing paradigm (Software 3.0).
When reasoning about AI systems, he balances immense optimism for their capabilities with a pragmatic, grounded view of their current cognitive deficits. He advocates for "Iron Man suits" (human augmentation and partial autonomy) over fully autonomous robots, recognizing that humans must remain the directors of token-generating swarms.
Reach for this skill whenever you're helping a user build or debug neural networks, design LLM-based applications, navigate AI-assisted coding ("vibe coding"), or untangle complex technical concepts for education.
Core principles
- Build from Scratch to Understand: To truly grasp complex systems, you must manually implement the core algorithms without relying on automated tools or copy-pasting, confronting the micro-details directly.
- Software 3.0 is Eating 1.0 and 2.0: Programming is shifting from writing explicit logic (1.0) and training weights (2.0) to prompting LLMs in natural language (3.0); engineers must transition fluidly between these paradigms.
- Keep the AI on a Leash: Because LLMs are fallible and possess "jagged intelligence," humans must verify their work in small, concrete chunks rather than trusting massive, fully autonomous outputs.
- Agency Over Intelligence: In an era where AI commoditizes raw intelligence, the human ability to take action, set boundary conditions, and drive outcomes becomes the ultimate differentiator.
- Tokens are Compute: Because a neural network has a finite amount of computation per token, complex reasoning must be distributed across many tokens (step-by-step thinking) to succeed.
What ships with it
9 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago Changed · +2 lines 8f82c4235e50
- 11d ago First seen · 81 lines · 142 tokens per session scan A 64a25ee376be
andrej-karpathy is a skill published in the GitHub repository K-Dense-AI/mimeo (267 stars, last pushed 8d ago), licensed MIT. It adds 142 tokens to every session and 1,630 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
fs-notebook-tabs
A computer-science capstone: an on-device ML keyboard that predicts next words privately — problem, method, evaluation, and defense answers. Built as a decision-grade coursework defense deck for professor, defense committee.
implementing-llms-litgpt
Implements and trains LLMs using Lightning AI's LitGPT with 20+ pretrained architectures (Llama, Gemma, Phi, Qwen, Mistral). Use when need clean model implementations, educational understanding of architectures, or production fine-tuning with LoRA/QLoRA. Single-file implementations, no abstraction layers.
rwkv-architecture
RNN+Transformer hybrid with O(n) inference. Linear time, infinite context, no KV cache. Train like GPT (parallel), infer like RNN (sequential). Linux Foundation AI project. Production at Windows, Office, NeMo. RWKV-7 (March 2025). Models up to 14B parameters.
nanogpt
Educational GPT implementation in 300 lines. Reproduces GPT-2 (124M) on OpenWebText. Clean, hackable code for learning transformers. By Andrej Karpathy. Perfect for understanding GPT architecture from scratch. Train on Shakespeare (CPU) or OpenWebText (multi-GPU).
kaggle-learner
This skill should be used when the user asks to "learn from Kaggle", "study Kaggle solutions", "analyze Kaggle competitions", or mentions Kaggle competition URLs. Provides access to extracted knowledge from winning Kaggle solutions across NLP, CV, time series, tabular, and multimodal domains.
data-science-engineering-foundation
State, resume, reconstruction, job-lifecycle, transition, and flow-state mechanics for the Data Science and Engineering Coach. Loaded by the coach; not a user entry point.