A curated library of the best AI learning resources—courses, tutorials, docs, and talks on agentic AI, MCP, Agent Skills, RAG, machine learning, fine-tuning, models, and frameworks, for AI engineers and technical teams.
A from-scratch blueprint for building, evaluating and securing agentic systems before you reach for a framework.
Build an agent harness from scratch, evaluate it rigorously, then train the model behind it with SFT and RL.
A free, tool-agnostic course on shipping real software with AI coding agents without losing engineering discipline.
One structured, open-source reference for serving LLMs fast, cheaply and reliably in production.
How DeepSeek got a 60-85% per-user generation speedup in production — the speculative decoding design, with training code and checkpoints released.
One Rust proxy in front of every agent hop — LLM provider, MCP tool server, agent-to-agent — with auth, rate limits, guardrails and OTel traces.
The resources learners opened most this week.
The MCP server that lets an agent drive a real browser from the accessibility tree instead of screenshots.
Turn an agent run into one OpenTelemetry trace — model calls, tool calls and the slow database query that actually caused the timeout.
Stanford's course on building software with coding agents: lectures, guest talks and assignments, all public
A free, battle-tested field manual for training and serving LLMs on real GPU clusters
Hugging Face's field guide to LLM benchmarks, score reproducibility and building your own evals
A frontier-scale MoE card that publishes its sparse-attention design, its serving flags and its own known failure modes.
The post-training stack that actually trained GLM — Megatron for training, SGLang for rollout, documented example by example.
The reference for the self-hosted gateway pattern — chat app in, agentic loop and skills out, running on your own box.
The only agent harness that documents its own loop, session log and tool pipeline as replaceable plugins.
Build restartable, scalable synthetic-data and AI-feedback pipelines instead of one-off generation scripts.
The manual for the largest open-source LLM app platform — visual agent workflows, a node-based RAG pipeline, and a self-hostable backend.
How Anthropic Labs built a planner/generator/evaluator harness that codes autonomously for hours — with the cost and wall-clock numbers.
The official manual for Google's open-source terminal coding agent — extensions, subagents, hooks, sandboxing and MCP.
Use the datastore already in your stack as the vector index, the agent's memory, and the semantic cache in front of your LLM bill.
An agent is one Python class: methods are capabilities, fields are state, docstrings are prompts, type hints are the contract.
Turn messy PDFs into clean Markdown locally, with open weights, open data, and a benchmark that actually checks tables.
Stop your agent relearning the same lesson every session — build skill induction, code knowledge graphs, and know when to fine-tune.
Write the CUDA kernels behind FlashAttention, paged attention and MoE, then benchmark your own GPT-2 training run.