Skip to main content

Labs

The twelve-week teaching schedule: what happens in each lab, what to prepare, and what leaves the room.

Every teaching week pairs its lecture with one weekly session: a formal lab in Weeks 2, 6, and 10, turning a central mechanism into a bounded, supported exercise; a drop-in clinic in Weeks 3, 7, and 11, where you bring evidence from your own project for diagnosis; or a guided session in the remaining six weeks, a short launch, validation, evaluation, or packaging activity that connects the lecture and the project without creating extra graded work.

Project 1 Launch and Data Validation

Verify the supported environment and data checksums, and write a bounded Project 1 target before selecting an architecture

Week 1 · 22 February 2027 · Guided session

Lab 1 — Spend a 32M Parameter Budget

Trace shapes through a causal decoder block and compare two legal architectures under the 32M/33.6M parameter boundary

Week 2 · 1 March 2027 · Formal lab

Project 1 Pilot Clinic

An evidence-led drop-in clinic for diagnosing a training pilot before committing to the principal Project 1 run

Week 3 · 8 March 2027 · Drop-in clinic

Project 1 Evaluation and Packaging

Run the released evaluation pipeline, verify a fresh-process load, and reconcile the report, ledger, and model package before submission

Week 4 · 15 March 2027 · Guided session

Project 2 Target and Baseline Planning

Define a target voice in observable dimensions, freeze the unchanged baseline, and cost one viable post-training route

Week 5 · 22 March 2027 · Guided session

Lab 2 — Build and Inspect a Post-Training Batch

Render a response-only loss mask, run a bounded SFT update, and test whether an apparent behaviour survives prompt paraphrase

Week 6 · 29 March 2027 · Formal lab

Project 2 Behaviour Clinic

An evidence-led drop-in clinic for interpreting target-versus-retention evidence and difficult preference labels

Week 7 · 12 April 2027 · Drop-in clinic

Project 2 Blind Evaluation and Packaging

Compare the unchanged and tuned checkpoints under blinded, matched conditions and package a reproducible Project 2 submission

Week 8 · 19 April 2027 · Guided session

Project 3 Task Contract and Baseline

Choose a permitted track and bounded target behaviour, write a valid task contract, and establish the unchanged starting-model baseline

Week 9 · 26 April 2027 · Guided session

Lab 3 — Train and Verify a Specialist

Turn a task contract into valid training and evaluation records, build a verifier before training, and test one shifted slice and one retained capability

Week 10 · 3 May 2027 · Formal lab

Project 3 Generalisation Clinic

An evidence-led drop-in clinic for locating the generalisation boundary and separating supported conclusions from alternative explanations

Week 11 · 10 May 2027 · Drop-in clinic

Project 3 Fresh-Load Audit and Submission

Reproduce the final specialist in a clean environment and submit one internally consistent evidence chain

Week 12 · 17 May 2027 · Guided session