Notable additions to this repo, newest first. Loosely follows the Keep a Changelog format.
frameworks/03-pre-pmf-validation.md— new "Lovers, likers, and the disinterested" section. Partition users by intensity, then build for the two bands that are already engaged — the ones who love the product and the ones who kind of like it — around what they have in common. Covers the three-band table, behavioral banding, the three-axis commonality (who they are, what they did before, what they do in the product), and three cuts of the feature data: the shared core (what both engaged bands use, chronically under-funded because it doesn't look like a differentiator), the separators (what lovers do that likers don't — simultaneously the depth investment and the conversion hypothesis for moving the middle up), and disinterested-only features as de-investment candidates. Plus "Pulling the middle up," seven traps, and the re-aiming decision when the lovers turn out to be a segment you never targeted. The "10 Happy Users" rule, the Sean Ellis section, the smile curve, and the phase's closing questions were extended to connect to it.decision-making/prioritization.md— new "Route requests by band before scoring them" section. Four roadmap lanes (hold the shared core / deepen the love / pull the middle up / explicitly not serving), plus a table of the six structural forces that pull the roadmap toward the users who never engaged, each with a counter-move.decision-making/finding-pmf.md— new "Iterate toward the lovers and the middle, not the disinterested" subsection under "Iterate on the who," giving the pivot a concrete direction once anything is live: the lovers define the target, the likers are the nearest growth, and what both share is the product.decision-making/customer-interviews.md— new "Interview the people who stayed, not only the ones who left" section: the "why do you stay" round run across both engaged bands, with different questions for each (what makes it indispensable vs. what's holding them at "kind of like it"), the substitution test, and referral history. Plus a matching anti-pattern.decision-making/metrics.md— Sean Ellis section extended: treat the survey as a segmenter, not a score. The three name lists are the deliverable; the percentage is not actionable. Applies to NPS and CSAT equally..claude/skills/pm-pmf-coach/SKILL.md— new step 5, "If anything is live, band the users and work the two engaged bands," with a matching output section and three new anti-patterns (roadmapping from churn, writing the middle off, ranking features by volume instead of by the three cuts)..claude/skills/pm-metrics-critic/SKILL.md— segment-decomposition step extended with the intensity check: a reported mean on any love metric has thrown away the data.
decision-making/agentic-product-design.md— new doc covering the design discipline for products that act rather than suggest. Why the deterministic-design assumption breaks, designing the constraint system instead of the screens, deciding the agent's persona before its scope, the inversion of the model-invisibility rule into explicit human-vs-agent attribution, calibrating transparency to who is accountable for the output, the three properties of visible failure (easy to see / fix / forgive), four trust mechanisms (inherited permissions, produce-vs-publish gate, stakes-based routing, autonomy earned on a measured override rate with deliberate over-gating at launch), familiar patterns, why the interface still matters (valued friction; chat as session log and guardrail rather than primary interaction; approvals next to the object they change), shipping to learn from real usage, and a 13-item agentic design checklist. Sourced from Figma's Writing the rules of agentic design (Expedia, Accor, Google, Capital One, Superhuman, Atlassian practice).decision-making/ai-integration.md— scope note distinguishing produce-then-user-acts AI from agentic AI, an inversion clause on Rule 4 (make the model invisible), an extension of the shape question for agents, and checklist item 10 routing agentic products to the new doc..claude/skills/pm-design-critic/SKILL.md— new step 3b, an agentic checklist (attribution, transparency calibration, failure visibility, produce-vs-publish, permissions, stakes routing, constraint spec, approval placement, valued friction) that runs when the design has AI taking actions rather than making suggestions..claude/skills/pm-framework-selector/SKILL.md— routing row for agentic product design questions.
decision-making/agentic-product-design.md— new "How many agents does the user see?" section on whether internal multi-agent topology should surface to the user. Internal decomposition is an engineering and governance decision; the user should still get one conversation and one resolution. Reconciles this with the attribution rule (attribute by system, not by internal boundary) and with named persistent agents. Checklist gains a topology-leakage item (now 14).
decision-making/first-principles-thinking.md— new framework doc. Five sections covering the narrow right use of first-principles reasoning (three regimes that justify it; default to the pattern otherwise), innovating on one or two axes (borrow the rest), the four hidden costs of throwing out an established pattern, combining first-principles with calling people who've solved the adjacent problem, and a decision checklist.- README "How to use this repo" table restructured by lifecycle phase (Frame & Decide / Discover & Validate / Build & Ship / Measure & Grow / Cross-functional / AI and the PM / PM craft and career / Rubrics, review, and operate).
- README "Reading paths" section consolidating the existing PMF path with two new paths: "If you're a working PM at a fast-growing company" and "If you're prepping for AI PM interviews."
- CHANGELOG.md (this file).
decision-making/activation.md— new "Day-1 / day-10 / day-30 sticky-moment tension" section. The sequencing problem most teams under-handle: day-1 features have to keep users alive long enough to discover the day-10 features that trigger word of mouth.decision-making/research-methods.md— two new sections: "Right-sizing exploration to stakes" (calibration table 3 vs 30 vs 300 users by stakes and reversibility) and "Map the full journey, not just your surface" (upstream/downstream questions).what-makes-a-good-pm.md— new "Career craft" section. Six sub-sections covering the graduated-ask networking pattern, the back-channel effect, breaking in via wedges and side doors, three interview signals (communication / agency / authentic enthusiasm), company-over-title for early career, earn-the-right-then-ask for internal moves.
decision-making/ai-craft-for-pms.md— new doc on how a working PM personally uses AI. Production-vs-thinking gap, eight prompt principles, anti-patterns table, AI/human decision matrix, hallucination types, four verification practices.templates/pm-prompt-library.md— new template with reusable prompts for common PM artifacts: discovery interview prep, transcript synthesis, PRD section-by-section, Amazon-style PR/FAQ working-backwards exercise, competitive monitoring setup and weekly synthesis.
decision-making/finding-pmf.md— new posture doc on finding product-market fit. Execution-doesn't-matter as the load-bearing claim, right-and-non-consensus, leap of faith vs. value hypothesis, need vs. desperation, iterate on the who not the what, savor the surprise, built to learn vs. built to execute, canonical distractions list..claude/skills/pm-pmf-coach/SKILL.md— new skill. Walks a founder or early-stage PM through finding PMF on a specific bet, end-to-end: surface leap of faith → build value hypothesis → design smallest falsifying experiment → interpret result → decide pivot/restart/savor → distractions check.- README "If you're finding PMF" reading-path section.
frameworks/06-product-process-for-pmf.md— new framework. The head-of-product decision framework. Eight-step spine, ten-principle counterintuitive doctrine, build-to-learn vs. build-to-execute culture, financing milestones mapped to PMF stages, decision checklist. Each step later expanded with why-of-each-step and skip-cost explanations.
ai-and-pm.md— two new sections: "Evals replace the PRD for model work" and "When to ship before the model is ready" (five-question framework).what-makes-a-good-pm.md— new pattern #8: "They stay personally close to the raw user signal." Existing AI-prototype pattern renumbered to #9.
decision-making/positioning-and-value-prop.md— new doc. The first-ten-touches alignment principle, the drop-in-but-worse reframe trap, killer demo as positioning North Star, three signals of broken positioning.pm-design-criticSKILL.md — new anti-pattern: "Missing the org pattern" for surfaces that look like the visible compromise between competing internal teams.frameworks/02-defining-the-mvp.md— added "Hack track vs. polished track for developer products."decision-making/prioritization.md— added "When user-demand ranking is wrong" section (override-axis sequencing: diagnose the bleed, map features to axes, apply three cross-cutting filters, sequence with parallelism).