What the lab knows
A Karpathy-pattern wiki the lab maintains across sessions — append-only by norm, cross-linked, read by humans and agents alike.
0 folders0 notes0 match tags
alleducation55library55openalice33m1328research24alice-core20design20architecture17inference16karpathy15market-intel12agents11reasoning11atlas10audit10inspector10serving10competitive9grpo9llm9weekly9alignment8foundations8rl8embed7fundamentals7kv-cache7lab7long-context7mcp7persona7quantization7rag7retrieval7training7transformer7transformers7academy6activity6attention6digest6rlvr6brand5deepseek5efficiency5evaluation5frontier5identity5infrastructure5multimodal5observability5positioning5ppo5reward-model5test-time-compute5tokenization5verifier5a2a4alice4benchmarks4bpe4generative-models4latency4llm-from-scratch4m114pricing4pytorch4rust4safety4strategy4agent-security3agentic3auth3autograd3backpropagation3best-of-n3canon3chain-of-thought3contamination3cost3deep-learning3diffusion3distillation3embeddings3engineering3fine-tuning3from-scratch3frontier-models3gpt-23gpu3grounding3image-generation3interpretability3linear-attention3memory3model-routing3moe3multi-agent3nanochat3onboarding3performance3post-training3product3rlhf3roadmap3rope3sampling3sequence-models3speculative-decoding3tiktoken3tool-use3agent-memory2ai-native2ai-safety2avatar2build2buildkit2byte-level-bpe2c2cache-mounts2capstone2cargo2ci2claude2claude-code2continuous-batching2conventions2credit-assignment2curriculum2dashboard2decisions2decode2deepseek-r12deltanet2demo2dispatch2embodied-ai2ensemble2eval-crisis2file-sd2flow-matching2fusion2generation2governance2gradient-descent2gtm2h2o2hetzner2induction-heads2inference-time2interoperability2json-schema2kernels2latent-diffusion2learning2llm-as-judge2llm-judge2long-horizon2lore2lto2mamba2mascot2mechanistic-interpretability2memory-bandwidth2minbpe2mla2moa2moat2model-based-rl2mold2multi-model2nanogpt2neural-network2nextest2o12on-device2path-deps2phase-12planning2prefill2primitives2prm2products2prometheus2prompt-injection2prompting2provenance2representation-learning2reward-hacking2rnn2routing2scaling2scaling-laws2sccache2security2self-consistency2sft2slm2snapkv2sora2sovereignty2ssm2streaming2sub-agents2swe-bench2systems2task-vectors2throughput2toolchain2training-loop2transformer-alternative2tts2unit-economics2vae2vector-search2video-generation2voice2voyager2vrm2vtuber2wonderland-facility2workspace2world2world-models2
infrastructure
1 noteagents
1 notelibrary
7 notes- Agentic RL & Long-Horizon Coding — training agents over multi-step trajectoriesstable2026-06-16libraryeducationagentic-rlrllong-horizon
- Autoencoders2026-06-16autoencodersrepresentation-learningself-supervised-learningdenoising-autoencodersparse-autoencoder
- GRPO & policy optimization — dropping the critic, letting the group be the baselinestable2026-06-16libraryeducationgrpoppodapo
- Reinforcement Learning Fundamentals — MDPs, value, policy, Q-learning, REINFORCE, PPOstable2026-06-16libraryeducationrlreinforcement-learningmdp
- RLHF, DPO & Alignment — teaching a model what we wantstable2026-06-16libraryeducationrlhfdpogrpo
- RLVR — Reinforcement Learning from Verifiable Rewardsstable2026-06-16libraryeducationrlvrrlgrpo
- Variational Autoencoders (VAEs)2026-06-16generative-modelsvaelatent-variableselbovq-vae