Study Plans

Study Plans

The repo's learning paths are organised by role. These are organised by calendar — the same material, arranged around a life that already has a job in it.

Pick the one whose time budget you can actually defend to yourself.

---

Which plan?

Four questions, six plans. The first question is the one people skip, and it is the one that changes the answer most: a plan for a team is not a faster plan for one person, it is a different plan.

_{▸ Open the picture full size}

Every plan ends in something you made, not something you read. That is the only property they share, and it is the one to keep if you invent your own.

One queue sits in front of four of them. Bedrock model access is granted per model and per region, on request, and approval is not instant. Start the account on night one of whichever plan you pick, even where the plan does not touch AWS until week two.

---

🧭 One week, no code · ~4 h — the method first

For a product manager, architect, delivery lead or sponsor who needs the method before the machinery. No AWS account, no Python.

Day	~45 min	Ends with
1	The Agentic PDLC and The Eight Loops	The eight loops marked closed, open or absent for your own programme
2	Your role page: PM · Architect · Engineering · QA · Sponsor	Your Monday list, five items
3	Gates & Governance and The Evidence Pack	The table for the P1 → P2 hard gate (the sign-off before anything is built), filled in honestly for one feature
4	Formulas & Calculators, then Sets A and B of Exercises	An acceptance bar you derived rather than guessed
5	Two scenarios from outside your industry, plus Anti-Patterns	Three anti-patterns you recognise, each with an owner
Any	One simulation in the workbench, from the live manual	A decision you got wrong once, cheaply

This is the plan to run before a programme starts, and the one to hand a sponsor who asks what they should read.

---

🌙 Four weeks, evenings · ~6 h/week

For a working engineer with an hour most evenings.

Week	Evenings	Do	Ends with
1	Mon–Thu	AWS setup (night 1, then wait for access) · Module 00 · PDL-01 lab	A classified use case of your own
2	Mon–Thu	Module 02 — onboarding notebook, Converse masterclass, exercises 1–2	A working tool-use round trip
3	Mon–Thu	AGL-01 → AGL-02 → AGL-03 · then Module 05	The loop, by hand, and then in Strands
4	Mon–Thu	Module 06 · TOOL-03 lab · your own agent	An agent in your domain

Start setup on night one even though you will not use it until week two. Model access approval is the only part with a queue.

---

🏃 One weekend · ~12 h

The Weekend Sprint, hour by hour. Do AWS setup the week before — a weekend spent waiting on model access is a wasted weekend.

	Sat AM	Sat PM	Sun AM	Sun PM
3 h each	Model answering: onboarding notebook, Converse masterclass	Tool use: refund agent, exercise 3	The loop by hand, then Strands	Your own domain; capstone if energy remains

---

📅 Twelve weeks, reading group · 90 min/week

For a team. One session a week, pre-work before, discussion during.

Wk	Pre-work	Discuss
1	Module 00	Bring one candidate use case each. Classify them together
2	Autonomy Ladder, Scope Fence	Which rung does ours need? What would we cut?
3	Module 02	Everyone runs the Converse notebook beforehand
4	Token Tax Ledger	Cost our use case. Argue about the assumptions
5	Module 05 + AGL labs	Everyone writes the loop by hand. No exceptions
6	Module 06	Design our tool catalogue together
7	Module 07 + Handoff Multiplier	Pick a topology and compute its H×
8	Module 10 core labs	Where would our corpus break retrieval?
9	Grounding Triangle	Read five of our own cited answers. Do the passages support them?
10	Module 13	Draft our golden set. Include cases we fail
11	Module 11	Who owns the corpus? Who is on call?
12	Scorecard	Score our system. Two people score independently and compare

The week-12 disagreement is the point. Where two scorers differ is your actual agenda.

---

🚪 No AWS account yet · ~20 h, £0

Everything here runs offline. Genuinely useful while access is pending — or if you never enable it.

Module 00 · Module 01 · Module 15	~13 h
All 10 L.A.B. labs — every one is offline and stdlib-only	~5 h
rag_by_hand.py · quality_gate.py	~2 h
The whole field guide	as long as you like

---

🎯 Interview in two weeks

Ruthless. These four labs plus two guides answer more interview questions than the rest combined.

Days	Do
1–3	AGL-01, AGL-02, AGL-03 — be able to write the loop from memory
4–5	EVAL-03 — and read quality_gate.py
6–8	Your role's interview guide — prepare the three stories it asks for
9–11	Autonomy Ladder, Evidence Ladder, Failure Signature Catalog
12–14	Build one small agent end to end. Measure its cost per resolved task. Bring the number

The four questions to be able to answer cold: what does it do when it does not know · how do you know it works, and on what evidence rung · what did you decide not to build · what did you get wrong and what changed.

---

Using these

Nobody finishes a plan exactly. The value is having a next action rather than a 16-module directory.

Share yours. If you invented a schedule that worked around your constraints, add it — this page is editable by anyone. Real plans from real people beat designed ones.