~/blog · filtered by category

pm.

11 entries filed under this signal. ← back to all entries

[blog/category/pm] · reverse chronological$ ls -la --filter=pm

2026-07-20pmTest the riskiest assumption before the agents build the whole thingWhen building was slow, you tested the riskiest assumption to avoid wasting months. Now agents build the whole thing before you have identified the assumption at all. That is the new trap.#riskiest-assumptions#product-management#ai-coding#experiments#paelladoc8 min./read →2026-07-20pmPrototype as spec: impossible to misinterpretA written spec is a description of the thing. A working prototype is the thing. One of them can be misread. Now that building is cheap, you should be shipping the one that can't.#prototype-as-spec#product-management#ai-coding#specs#paelladoc8 min./read →2026-07-20pmProduct discovery with AI: partner for thinking, not for decidingAI is an excellent thinking partner in discovery and a terrible deciding partner. The line between the two is where good product work now lives, and it is easy to cross without noticing.#ai-discovery#product-management#ai-coding#product-research#paelladoc8 min./read →2026-07-20pmDecision logs: ending the "I never agreed to that" problemHalf the arguments on a product team are not about the future. They are about what was decided in the past, and nobody wrote it down. A decision log is the cheapest fix you are not using.#decision-logs#product-management#ai-coding#product-memory#paelladoc9 min./read →2026-07-20pmYou write PRDs nobody reads. The artifact is the problem.The PRD is not failing because you write it badly. It is failing because a document that is finished the moment it is written cannot keep up with a product that changes every day.#prds-nobody-reads#product-management#ai-coding#living-specs#paelladoc8 min./read →2026-07-20pmThe PM–engineering bridge: one system from discovery to merged codePMs live in one content world and engineers live in another. The handoff between them is a translation loss nobody owns. This is about closing the gap almost no team crosses.#product-management#engineering#ai-coding#discovery#paelladoc9 min./read →2026-07-20pmInsight → decision → outcome: the chain AI speed keeps breakingA feature exists because someone learned something, decided something, and shipped it. When agents collapse the middle, that chain snaps and nobody can walk it back.#product-management#ai-coding#decision-making#product-outcomes#paelladoc8 min./read →2026-07-20pmProduct decisions with evidence attachedEngineering learned that a claim of completion has to travel with its proof. Product decisions almost never do. The reasoning floats free of the decision, and by next quarter nobody can audit it.#product-management#decision-making#evidence#ai-coding#paelladoc7 min./read →2026-07-20pmAI-assisted user research without laundering your evidenceAI is genuinely good at synthesizing forty interviews into themes. It is not a substitute for the interviews. The line between the two is where research quietly turns into fiction.#product-management#user-research#ai-coding#discovery#paelladoc8 min./read →2026-07-20pmAI slop with a nice format: the failure mode of AI-assisted product workAI did not make your product team decide better. It made you produce more PRDs, more tickets, more decks, faster. More output, same decisions. That is the failure mode, and it looks like productivity.#ai-slop-product#product-management#ai-coding#decision-quality#paelladoc8 min./read →2026-07-20pmAn operating model for product teams that build with agentsAn operating model is roles, artifacts and rhythm. When agents do the construction, all three change. Most teams kept the old three and wonder why speed turned into chaos.#product-management#operating-model#ai-coding#team-process#paelladoc9 min./read →

— end of log · 11 entries in pm —