NEWThe AI Product Builder's Playbook is out. 17 chapters, free to read.The Playbook is out. Read it free.→

Work with me Playbook Off Market Articles

All parts

00The role: the AI Product Builder
01The laws of building
02The dev foundations (plain-language)
03CI/CD and mechanical checks
04The work cycles
05Specs and scoping (the core craft)
06Tests and QA: proof against intention
07Data, migrations, and money
08Shipping, security, and operations
09Documentation as infrastructure
10Design system, content, coherence
11Building with AI: what actually changes
12The catalogue of traps
13Going further (what I haven't done yet, and what dev+++ looks like)
14Glossary
15The self-assessment
16Case notes (anonymized)

Part 15

The self-assessment

2 min read

In short

Twenty questions to place yourself. Count the yeses.
Process, shipping, proof, data, security, and how you work with AI.
A read becomes a tool the moment you score yourself.

0out of 20

Shipping on hope

you're shipping on hope; start with the push rule, the early test, and a real definition of done.

Process and decisions

Do you write a short spec before the build, not after?
Do you decide one trade-off at a time, in writing, with the rejected option noted?
Can you say, right now, what "done" means for you, and does it include "in production and seen working"?
Do you slice vertically, a thin complete journey first, then depth?

Version control and shipping

Do you push after every unit of work, always?
Before you deploy something risky, do you know exactly how you'd roll it back?
Do you keep a changelog and version your releases?
Do risky paths (money, webhooks) sit behind a switch you can flip without a redeploy?

Proof over intention

Have you seen each of your safety checks fail on purpose at least once?
Do you test the real journey early, from the first slice, not at the end?
On money paths, do you validate with a real test event, not just a code read?
When you change data at scale, do you prove it with an invariant, not a re-read?

Security, data, ops

Is your full-access key server-side only, never in the frontend?
Do you know where every environment variable lives and who sets it?
Is personal data collected minimally, protected, and deletable on request?
Do you learn about a production problem from a dashboard, or from an angry user?

Documentation and AI

Could a cold newcomer, human or agent, get productive from your docs alone?
Are your specs written so an agent can't reasonably misread them (explicit scope, do-not-touch, one goal)?
Do you keep a project memory with a budget, one that tends toward empty?
When it ships, are you clear that the decision, the verification, and the accountability were yours?

Where to next

Download the PDFPDF · 1.1 MB

I'm a…

buildercompany hiring

Your email

Website

← PreviousGlossary Next →Case notes (anonymized)

Benoit Lecureur · Paris

LinkedIn · Malt