Understand AI.
Build your understanding, one experiment at a time.
A reading and experiment space for AI engineers and researchers. Study the ideas, follow a learning path, and test what you learn.
A complete learning route
Read → implement → evaluate
Frontier research
Eight selected reads. From research ideas to experiments you can build.
Featured readRobot recovery: when retrying helps, and when it hides failure
Posted
Read Backtrack, DexPort and ViT-H through a practical lens: frozen policies, action labels, executable motion and bounded recovery. Includes a local state-machine experiment.
Research notesSelf-improving agents: reward gains are not capability gains
Posted
A research reading guide and local experiment on adaptive harnesses, reward collapse, frozen-policy comparisons and held-out evaluation.
Research notesFrom LLM Calls to Agent Harnesses
Posted
An English reading brief and hands-on tutorial on agent runtimes: context, tool permissions, checkpoints, bounded execution and reproducible failure tests.
Research notesTest-time compute: search, verification and budgets
Posted
Prerequisites: probability, evaluation and inference cost. Learn when more attempts help and when a verifier becomes the bottleneck.
Research notesSelective state spaces: Mamba and sequence memory
Posted
Prerequisites: linear algebra, recurrence and attention. Compare compressed state with explicit token retrieval.
Research notesDiffusion language models: masking and iterative decoding
Posted
Prerequisites: cross-entropy, Transformers and generative modeling. Understand LLaDA-style masked generation without equating parallelism with speed.
Paper analysisDeepSeek-R1: rewards, reasoning and what the evidence establishes
Posted
Study R1-Zero and R1 separately, then build a controlled test of reasoning improvement. This analysis is pinned to arXiv v1.
Paper analysisLLM-as-a-judge: evaluate the evaluator
Posted
Separate answer preference from correctness and reveal how order effects can alter a reported win rate.
Learning library
Choose a route through the material.
AI / LM curriculum
From foundations to language models.
Reinforcement learning
Value functions, policy optimization and reasoning.
Physical AI
Perception, control and robot learning.
Paper library
Mechanisms, derivations and evidence.
Latest tutorials
Recently added to the learning library
Write Multi-Head Attention from scratch
Trace Q, K and V, split and merge heads, then prove your causal mask works.
Read tutorial ↗Softmax that stays stable
Derive the gradient, work through a small example and avoid overflow.
Read tutorial ↗K-Means beyond the happy path
Implement assignment and updates; handle empty clusters and convergence.
Read tutorial ↗Adam, one update at a time
Understand persistent moments, bias correction and the first parameter update.
Read tutorial ↗Beam Search without hidden mistakes
Keep global top-k candidates, track EOS and compare against greedy decoding.
Read tutorial ↗Experiment locally
Open a tool when you have something to test.
Local tools process inputs in your browser. No API key or input uploads.
Interview & projects
Practice with evidence, not memorized answers.
Interview practice
Questions with answers and worked reasoning.
Projects & code
Reference implementations and acceptance tests.
Latest practice
Recent interview drills and hands-on projects
Handwrite Beam Search, then defend every choice
Practice global top-k selection, EOS handling and length bias with a worked example and reference code.
Read exercise ↗INTERVIEW / OPTIMIZATIONDerive Adam and trace the first update
A paper-and-pencil drill in moving averages and bias correction, followed by a tested implementation.
Read exercise ↗PROJECT / TRANSFORMERSProve your KV cache is correct
Build causal attention and compare full-sequence outputs with incremental decoding using numerical invariants.
Read exercise ↗PROJECT / RAGBuild retrieval with a measured baseline
Measure retrieval accuracy, answerable recall and correct abstention separately; inspect the failure cases.
Read exercise ↗