Learn by doing
Make something real
Agentic product management: the whole product discipline — decide, build, prove, grow — run by one person and their agents, on rails that keep the evidence honest. The Golden Frijoles methodology is how that is actually done: bring your own project, use your own agents, and learn the method while you make.
Direction · your North Star surrounds the loop
- ConsiderBring an idea → Design it → Place the Bet
- OperateBuild ↔ Prove
- ExitReconsider → Learn
Before you begin
You should already have:
- a Golden Frijoles project
- the full Direction produced by your North Star workshop
- a connected agent
- Golden Frijoles project rails available where you intend to work
If you completed first-run setup, you have this.
Do this on your project
Open the real project you want to change. Do not create a tutorial project.
Look for
Your project agent should be able to orient to project agents and instructions, ways of working, relevant learnings, team memory where available, and the Golden Frijoles context it has permission to read.
If not, finish setup before continuing.
Methodology index
Six chapters. One real project
Read in order the first time. Jump straight to a chapter when you come back.
Consider
01Bring an idea
Start with something you genuinely want to make, fix, change, or learn. An idea is not yet a commitment.
Read chapterConsider
02Design it
Orient to reality, establish the Outcome and Baseline, fix the Appetite, and find a plausible bounded approach.
Read chapterConsider
03Place the Bet
Make the investment decision explicit: Evidence, Displacement, Boundaries, Authority, and whether this is actually worth doing.
Read chapterOperate
04Build it
Route the work, Slice for coherent behavior, and Bound autonomy so humans and agents can operate without losing control.
Read chapterOperate
05Prove it
Separate deterministic verification, independent judgment, and Evidence from reality. Shipping is not proof of Outcome.
Read chapterExit
06Decide what happens next
Reconsider the Bet from Evidence, then turn Learning into an actual change in how the system behaves.
Read chapterYou completed the loop
You started with a rough idea.
You Considered whether it deserved investment.
You Operated a bounded human-agent system to make it real and establish Evidence.
You Exited the investment according to what reality justified.
You do not need to memorize a list of practices to do it again.
Your North Star provides Direction around the loop.
The deeper vocabulary is there when you need precision. The loop is there when you need to make something.
Practitioner checkpoint
The guide succeeds if you can now:
- turn a raw ask into a bounded investment decision
- set Appetite before allowing a solution to expand
- distinguish Outcome from Output
- displace something when prioritizing something else
- route work according to uncertainty and Authority
- grant autonomy inside explicit Boundaries
- build coherent Slices rather than technical task piles
- demand Evidence appropriate to the claim
- distinguish shipping from proving
- reconsider without sunk-cost loyalty
- turn Learning into a change in the operating system
Terminology recall is not the test.
Better operation is.