Showed that fine-tuning a GPT model on public GitHub code yields a capable program synthesizer, and introduced HumanEval — the docstring-to-function benchmark that still anchors code-generation evaluation. A production variant powers GitHub Copilot.
AI Infra2021
Runs, manages, and scales AI workloads across 20+ clouds, Kubernetes, Slurm, and on-prem from one YAML or Python spec. Auto-provisions GPUs/TPUs, fails over across regions and providers when capacity is short, and routes jobs to the cheapest option.
AI Client2021
Self-hostable personal “AI second brain” that turns web pages and documents into a searchable knowledge base, builds custom agents and automations, and connects to local or cloud LLMs with multi-platform access.
AI Infra2022
A continuously-updated, categorized self-hosting guide that catalogs tools, deployment notes and resources for containers, networking, home automation and running LLMs/chatbots locally; provided as a long, structured README with links and quick setup tips.
AI Deploy2022
Deploys PyTorch models directly on phones, microcontrollers, and embedded hardware via ahead-of-time compilation to a ~50KB C++ runtime. Delegates subgraphs to 12+ backends (XNNPACK, CoreML, Qualcomm, ARM Ethos-U) with torchao quantization.
AI Dataset2022
A 57-subject multiple-choice benchmark for measuring broad language understanding in LLMs; provides per-subject configs and test/dev/auxiliary_train splits for few-/zero-shot evaluation, widely used for model comparison and academic reporting.
AI Dataset2022
Provides 115M public GitHub source files (≈873GB of code, ~1TB uncompressed) with per-file metadata (repo, path, language, license). Supports streaming, language/license filtering and full download for training and evaluating code LLMs and code generation models.
AI Dataset2022
Provides cleaned, per-language snapshots of Wikipedia articles (id, url, title, text) packaged as Hugging Face dataset configs (Parquet). Covers 300+ language configs and dated dumps — useful for language modeling, multilingual NLP, retrieval, and RAG pipelines.
Made reinforcement learning from human feedback (RLHF) the standard alignment recipe: collect demonstrations and preference rankings, train a reward model, then optimize with PPO. A 1.3B aligned model was preferred over the 175B GPT-3 by human raters.
Official collection of example notebooks and guides for building with the OpenAI API — text generation, embeddings, function calling, RAG, fine-tuning, and more. Mostly runnable Jupyter notebooks (~93%); mirrored at cookbook.openai.com.
AI Agent2022
Terminal rebuilt around AI agents: orchestrate Claude Code, Codex, and Warp's own agent in parallel, each with codebase indexing and scoped permissions. Run them locally or in the cloud, and bring your own model via Bedrock, LiteLLM, OpenRouter.
AI Dataset2022
Benchmark dataset of ~8.5k grade-school math word problems with step-by-step solutions and calculator annotations for evaluating multi-step arithmetic reasoning in language models. Provided in two configs (main and socratic) and commonly used for chain-of-thought prompting, fine-tuning, and verifier training.