Cohort Session 4 Methods Decoded - akash-coded/aws-bedrock-agentcore-strands GitHub Wiki

Session 4 · Methods decoded

Where do AI-DLC, AIDD, BMAD and spec-driven development actually sit, and how much process does one change need?

Part of the Cohort Kit. Ninety minutes. The session that ends the method wars: by the end the room can place any named method on the four phases, has mapped its own current practice onto it, and has a rule for how deep to go on a change.

Pre-reading (about 35 minutes; the four method lessons are optional depth)

AI-DLC vs AIDD vs agentic SDLC — the comparison, and why they are not competitors
One lifecycle for every method — how each method occupies part of the same lifecycle
How much process a change needs — depth of change as the decision, not method choice
Optional, one each, split across the room: What is AI-DLC? · What is AIDD? · What is the BMAD Method? · What is spec-driven development?

Bring. One page describing how the team delivers today: the ceremonies, the artefacts, the approvals. It does not need a name.

You will leave with. The team's current practice placed on the four phases, the parts of any named method it will borrow, and a written rule for depth of change.

Run-sheet

Minutes	Block	Do this	Material
0–10	Recap	Which method owns P0? Which is a build-phase discipline? What is the one decision that matters more than method choice? Anyone who read an optional lesson gives its one-sentence summary	—
10–35	The idea	The four-methods picture on screen, then the workbench's compare page with two methods the room argues about most. Then the frameworks page's lineage table, so the room sees which claims are documented and which are the manual's own	Four methods, one lifecycle · Compare · Every framework, with its lineage
35–65	Exercise	Pairs place the team's own practice on the four phases: which phase each ceremony serves, which artefact each approval produces, which loops it closes. Then each pair names one thing it would borrow from one named method and says which gap it fills	The "bring" page · The Agentic PDLC · Decision tree 4 · hard gate or soft gate?
65–80	Decision	The room writes its depth-of-change rule: three tiers of change, what each needs before it ships, and one example of each from the candidate project. Then classifies the first three bolts from session 3 against it	How much process a change needs · Hard or soft gate classifier
80–90	Close	Homework named. One sentence each: which ceremony does the team run that closes no loop?	—

Facilitator notes

This session has no arithmetic and can drift into opinion. The four phases are the anchor: every claim about a method has to be placed on them. "AI-DLC is better" is not placeable; "AI-DLC's mob elaboration is our P1 spec session" is.
The lineage table matters. Some of what circulates about these methods is vendor material. The frameworks page marks confidence per framework; show the room how to read the mark.
Depth of change is the takeaway. A team that leaves with a three-tier rule and examples has the whole session. A team that leaves having chosen a method has the wrong thing.
Ten-minute version: the four-methods picture, then straight to the depth-of-change rule.

Homework

One page: "how we deliver", the team's practice on the four phases with the borrowed parts marked, and the depth-of-change rule at the bottom. This becomes the first page of the team's own operating guide; the cadence lesson in session 8 adds the rhythm.