Data Science Origin
From first principles to the frontier

One path through modern AI.

Twenty articles, ordered as a single climb — start knowing nothing, finish able to build agents, train reasoning models, and ship. No prerequisites. Nothing to pay.

Start the pathBrowse articles
20Articles
5Tiers, beginner → frontier
$0Free, no account
Jul 26Kept current
How it's built

Five tiers, one climb

Every article assumes only the ones before it. You don't pick a course — you follow the rungs. Land on tier one knowing nothing; leave tier five able to build.

FoundationsWhat actually happens inside an LLM — tokens, attention, reasoning models.
3 articles
Core skillsContext engineering, MCP, RAG, memory, embeddings — the working toolkit.
5 articles
Building agentsThe agent loop, orchestration, evals as governors, harness engineering, security.
5 articles
Models & trainingRLVR, fine-tune vs prompt, MoE & hybrid architectures, the cost collapse.
4 articles
FrontierSmall models beating giants, running locally, world models & physical AI.
3 articles
Start reading

Fresh off the path

World models and physical AI: what comes after chat
Advanced

World models and physical AI: what comes after chat

From predicting text to predicting the physical world — what world models are and why robotics needs them.

Frontier · 13 min
Running frontier models locally: what it takes in 2026
Intermediate

Running frontier models locally: what it takes in 2026

The hardware and quantization tradeoffs for running capable models on your own machine.

Frontier · 11 min
Small models beating giants: the efficiency frontier
Advanced

Small models beating giants: the efficiency frontier

How distillation and better data let small models match frontier models on narrow tasks.

Frontier · 12 min

One email when a new rung ships

No digests, no noise. You get a note the day a new article joins the path — and nothing else.