The curriculum
Study
The syllabus, published before it is climbed. Every unit runs consume → do → output, and nothing closes without an artifact: a notebook entry, a lab build, a playable explanation, or a spaced-repetition bottle.
LADDER SET · LAST PASS
On the bench
Written by the study desk, not computed hereBuild 01 — predict this laptop’s decode rate from memory bandwidth, then measure it.
Under docs/brand/learning-doctrine.md the bench leads with what gets built, not what gets watched. Build 01 is the lowest-order open build, it runs on hardware already verified, and it is the cheapest possible test of the doctrine itself.
Track A — the AI-infrastructure stack
Bottom-up · rungs gate each otherA bottom-up ladder from the machine to the money: what the hardware physically does, what it costs to serve a token, and what that arithmetic implies about the companies selling it. Rungs gate each other — a rung is not opened until the one under it can be explained from memory.
| Rung | Unit | State | Progress | Artifacts | Seat |
|---|---|---|---|---|---|
| 1 | Digital Design & Computer ArchitectureSAFARI · ETH ZürichConditional — Studied only if the Computer Architecture L1–L2 gate-check fails. Connor decides, honestly; the desk records the verdict. | Conditional | Not scoped — it is only sequenced if the gate-check fails. | NBLABGMEBTL0/1 | Read it LINK PENDING |
| 2 | Computer ArchitectureOnur Mutlu · ETH Zürich · Fall 2023 | Ready | 0 / 32 lectures | NBLABGMEBTL0/2 | Read it ↗ |
| 3 | Inference EngineeringPhilip Kiely · Baseten · 2026Gated — Opens when the memory rung (L2a, L9, L12–L15) closes — the hardware has to be under it. | Gated | 0 / 8 chapters | NBLABGMEBTL0/3 | Read it ↗ |
| 3 | Efficiency in LLMsAlex Smola · companion to rung 3Gated — Shares rung 3 with Inference Engineering and opens with it. | Gated | 0 / 6 sections | NBLABGMEBTL0/1 | Read it LINK PENDING |
| 3 | GPU Kernels / MLSysMLSys kernel trackGated — Needs the memory hierarchy under it before occupancy and coalescing mean anything. | Gated | 0 / 4 parts | NBLABGMEBTL0/1 | Read it LINK PENDING |
| 4 | CS336 — Language Models from ScratchStanfordGated — Follows rung 3. A PyTorch readiness check precedes Assignment 1; if fluency is thin, a nanoGPT warm-up comes first. | Gated | 0 / 19 lectures | NBLABGMEBTL0/2 | Read it ↗ |
| 5 | Agentic AIAgentic AI trackGated — Follows CS336. The deep-RL family only after the CS336 alignment on-ramp is logged. | Gated | Syllabus length not fixed — no published lecture count to cite. | NBLABGMEBTL0/2 | Read it LINK PENDING |
| — | RLHF & Post-TrainingNathan LambertNo quota — No weekly slot allocated. Draft Day decision. | Awaiting Connor | 0 / 13 video lectures | NBLABGMEBTL0/1 | Read it ↗ |
| — | Build a Reasoning Model From ScratchSebastian Raschka | Ready | 0 / 8 chapters | NBLABGMEBTL0/3 | Read it ↗ |
Valuation and systematic risk, run as a daily loop rather than a course. It is read here so the two tracks are one picture, but the work happens on the markets desk and every action hands off there — this surface never re-implements the cadence.
| Rung | Unit | State | Progress | Artifacts | Seat |
|---|---|---|---|---|---|
| — | ValuationAswath Damodaran · 12-week climb · case company Micron (MU)Awaiting — Module 0 cannot start until F1 is answered. The block is Connor’s, not the machine’s. | Awaiting Connor | 0 / 9 modules | NBLABGMEBTL0/2 | Read it ↗ |
| — | Systematic TradingRobert Carver · phases C0–C4Honest read — Code-advanced but mastery-dormant: the simulators built from this material pass 35 of 35 tests, and C0 is unstarted. The instrument runs; the operator has not been trained on it. | Ready | 0 / 5 phases | NBLABGMEBTL0/1 | Read it LINK PENDING |
| — | Smart PortfoliosRobert Carver · 10 bitesAwaiting — Cadence pending Connor’s Decision Q1. The desk surfaces the gap and never invents a default.No quota — No cadence set. | Awaiting Connor | 0 / 10 bites | NBLABGMEBTL0/1 | Read it LINK PENDING |
- BTLPipeline and cache vocabulary, on a spaced scheduleDigital Design & Computer Architecture →
- NBThe memory wall, derived rather than restated from a vendor deckComputer Architecture →
- BTLMemory-hierarchy numbers, on a spaced scheduleComputer Architecture →
- NBWhy memory bandwidth is the bottleneck, from the physicsComputer Architecture · L15 →
- NBThe token-price → task-price bridgeInference Engineering →
- LABThe memory roofline, verifiedInference Engineering →
- BTLops:byte and KV-cache formulas, spacedInference Engineering →
- NBServing unit economics — where the cost of a token goesInference Engineering · ch8 →
- LABVRAM / KV-cache calculator, verified against a real runInference Engineering · ch8 →
- BTLEfficiency levers and what each one costs, spacedEfficiency in LLMs →
- BTLOccupancy and coalescing rules, spacedGPU Kernels / MLSys →
- NBI built a language model from scratch — the build, in publicCS336 — Language Models from Scratch →
- LABThe from-scratch model, runningCS336 — Language Models from Scratch →
- NBThe harness flagship — the earned version of the argumentAgentic AI →
- LABDesign notes into the machine that runs this siteAgentic AI →
- BTLPost-training objectives and their failure modes, spacedRLHF & Post-Training →
- NBReasoning models, measured on one small enough to watchBuild a Reasoning Model From Scratch →
- GMEA self-refinement loop, playableBuild a Reasoning Model From Scratch →
- BTLGRPO mechanics and where it breaks, spacedBuild a Reasoning Model From Scratch →
- NBWhat test-time compute actually buys, measuredBuild a Reasoning Model From Scratch · ch4 →
- NBThe Micron valuation, with its assumptions exposedValuation →
- BTLValuation fact set, spacedValuation →
- BTLVolatility targeting and risk decomposition, spacedSystematic Trading →
- BTLHandcrafting and rebalancing rules, spacedSmart Portfolios →
Every slot above is currently owed, which is what a syllabus published before the climb looks like. Earmarked against workshop builds: 3 of the slots above, 3 of them against builds already underway. None is counted here until it ships there — the workshop is the authority on whether a build landed, not this page. Debt accrues on coverage, not on existence: an all-zero ladder is a contract written in advance, not a pile of failures.
Ledger findings
Targets named but not found on the owning surfaceNo unresolvable targets — every earmarked build and every landed artifact points at something that exists.
Candidates
Assessed 2026-08-10–2026-08-18 · not allocatedInference Algorithms for Language Modeling (11-664/763)
Graham Neubig · CMU · Fall 2025
free · Lectures free on YouTube; slides and code public.
- barely overlaps Inference Engineering — Neubig teaches inference ALGORITHMS (decoding, search); Kiely teaches inference SYSTEMS (serving).
- substantially overlaps Build a Reasoning Model From Scratch — Chain-of-thought, self-refinement and reasoning models are covered by both.
Cherry-pick the classical search lectures (beam search, A*, best-first) — nothing else on the ladder covers them. Watch the CoT / self-refine / reasoning lectures AFTER the matching Raschka chapters, as retrieval practice rather than first intake.
Building a Coding Agent From Scratch
decodingai-magazine · GitHub
free · 8 lessons, free, runs on Modal.
- substantially overlaps Agentic AI — Harness architecture end to end: agent loop, durable execution, sandboxing, context engineering, subagents, evals, remote swarms. A candidate feeder for rung 5, not a substitute for it.
The natural feeder for the-machine once rung 5 is live. Not a substitute for the rung, and not startable ahead of it without inverting the ladder.
Harness Engineering for Self-Improvement
Lilian Weng · 2026-07-04
free · A single essay, roughly 31 minutes.
- barely overlaps Agentic AI — One essay, directly on the harness thesis — vocabulary, not coverage.
Read as a primer before rung 5 opens. Cheapest possible entry to the harness vocabulary; not a unit.
CS329A — Self-Improving AI Agents
Aakanksha Chowdhery & Azalia Mirhoseini · Stanford · Autumn 2025
free · 20 sessions taught, 3 homeworks + a final project; a condensed lecture series is public.
- substantially overlaps Build a Reasoning Model From Scratch — Near chapter-for-chapter: ch3↔Robust Verification, ch4↔Test-Time Compute Scaling, ch5↔Learning from Feedback, ch6–7↔Train-Time Scaling/Scaling RL. Raschka builds it small; this is what a lab found building it large.
- substantially overlaps Agentic AI — Rung-5 territory taught with mechanism rather than survey — memory, long-horizon tasks, agentic evals.
The one item here worth acting on immediately, and NOT as a unit — as the retrieval half of reasoning-from-scratch, which is the only open, ungated, environment-verified unit on the ladder. It rides inside that existing quota and costs no new slot. Same rule as neubig-lm-inference: the matching Raschka chapter runs on the laptop FIRST, the lecture is watched after. Reversed, it yields a vocabulary that cannot be defended — the exact failure the seat test exists to catch.
Agentic AI MOOC
Dawn Song · UC Berkeley RDI · Fall 2025
free · 12 guest lectures, one industry speaker each; slides and recordings public. Optional AgentX competition.
- barely overlaps Agentic AI — Covers the rung’s territory as landscape, not mechanism — twelve practitioner talks. Vocabulary and a map of who is doing what; not the RL/RLHF/PPO/DPO/GRPO machinery the rung is for.
File as a primer, and use it to force a decision rather than to answer one. The agentic-ai unit currently carries NO source url and no lecture count, while docs/context/study-plan.md attributes rung 5 to Roitman — so rung 5 has one slot and more than one candidate meaning. Standing recommendation (Connor, 2026-08-18): Roitman remains the rung; this is landscape reading, watchable at any point. On the teaching ladder it is L1/L2 and earns a `read it` seat unless the AgentX project is actually entered.
CS 185/285 — Deep Reinforcement Learning
Sergey Levine · UC Berkeley · Spring 2026
free · 25 lectures, 5 homeworks + a final project. Slides and homework public; video links point at the Fall 2023 recordings.
- barely overlaps Build a Reasoning Model From Scratch — Raschka ch6–7 (GRPO) touch the same policy-gradient family, far more cheaply and already on the ladder.
- barely overlaps Agentic AI — Only the LLM RL lecture pair and its homework speak to the harness argument; the other 23 lectures are control and robotics.
The heaviest item on the list and the only one with a HARD published prerequisite Connor does not meet on the record: CS189/CS289 or equivalent, plus assumed working ability to train deep neural networks. CS336 (rung 4) is what clears that — and CS336 sits behind rung 3, which sits behind rung 2 at 0/32, so this is a genuine three-rung dependency. Recorded here with the prerequisite stated so it stops being rediscovered as "should I start this yet?" every few months. If only one slice is ever wanted, it is the LLM RL lecture pair + HW4; lectures 1–3 (imitation, behavioural cloning) are the approachable on-ramp.
Building a distributed training framework from first principles
Umar Jamil
free · One continuous first-principles build. Duration and publication date unverified.
UNWATCHED — title and channel confirmed via oEmbed, but the video itself has not been opened, so no earned overlap read exists. On its stated syllabus it is rung-4 material sitting directly on CS336 Unit 2, and it is the clean complement to gpu-kernels: that bridge is INTRA-GPU (occupancy, coalescing, tensor cores), this is INTER-GPU (process groups, collectives, device meshes, and pipeline/data/FSDP/tensor/context/expert parallelism composed into one framework). Neither covers the other. Its rare shape is the draw — one self-contained build with a working artifact at the end, where everything else at rung 4 is a 19-lecture course. The catch: "no prior knowledge of distributed training required" is true of the distributed part only; it also builds MLA, RoPE, YaRN and a mixture-of-experts layer, so without transformer fluency it is watched rather than built — a `read it` seat in a `ran it` costume. Revisit when CS336 opens, or earlier if the transformer fluency arrives ahead of the ladder.
Prime Agent
Prime Intellect · open source
free
UNVERIFIED — the code has not been opened, so no overlap read exists. Read the code before it is judged. Listed so the gap is on the record rather than rediscovered monthly.
CMU Agents course
cmu-agents.com
unreleased
Not publicly released. Tracked so it is not rediscovered every month.
- §1Architecture and Model Design22 papers · ★1
- §2Efficient Training and Scaling15 papers · ★1
- §3Inference Efficiency and KV Cache8 papers · ★1
- §4Sparse Attention and Long Context9 papers · ★1
- §5Reasoning and Test-Time Compute15 papers · ★1
- §6Reinforcement Learning and RLVR25 papers · ★1ModelsReasoning book ch6–8 · RLHF book
- §7Agent Systems and Tool Use33 papers · ★1Harnessesthe-machine
- §8Coding Agents and Software Engineering16 papers · ★1Harnessesthe-machine
- §9Diffusion Language Models9 papers · ★1
- §10Model Evaluation and Benchmarks13 papers · ★1
Presence on this list is a pointer, not a claim: a source with no close-read file has not been read, and counting it would inflate the shelf the way a timestamp inflates freshness. The catalogue itself lives in the repo at docs/context/readings/llm-papers-2026-h1.md — a working file, not a public route. It is a bibliography, distinct from docs/context/reading-list.json, which is the intake queue where a bookmark gets exactly one disposition — and now has a page of its own. The source list →
Where a saved link goes before it becomes anything. Each one gets exactly one ruling — read it, propose it for the Atlas, write something, build something, or bin it. The bins are on the page too; a list of only the keepers would flatter the judgement.
The curriculum is one typed file, and the study desk is its only writer. There is deliberately no tick-box on this page and there never will be — a second way to mark something done is a second version of the truth, and it would diverge within a week. The spaced-repetition bottles live off this site and are counted here, never shown. Track C was removed 2026-08-10; two tracks is the whole ladder. The workshop →