Every search term, and what it brought in

Round 1 search terms, committed before any data was pulled. They cover each stage of a product team's work (plan, spec, design, build, review, QA, release, feedback, docs, coordination) and are phrased neutrally. Each term casts a wide net: a relevance pass later removes posts that aren't about work pain. Blast Radius gets no terms of its own. 'waiting on engineering' covers it the same way it covers every other waiting pain.

Round 2 widens round 1. Round 1 combined quoted phrases with extra required words, which cut results sharply (e.g. '"code review" slow' found 209 posts, while '"code review"' alone finds 6,060). It kept 1,623 posts in total. Round 2 searches the core phrases on their own and adds common variants, over the same topics and the same window. This brings in more posts that aren't about pain at all. The relevance pass is there to remove them. No new topics, and still no terms for any candidate product.

Round 3 adds the views of designers and project/product managers, since Hacker News is mostly engineers (codebook v1 found non-engineers depending on engineering once in 300 HN posts, which HN alone can't settle). Sources: UX Stack Exchange and Project Management Stack Exchange. Both sites are small and entirely on topic, so the round takes every question and answer in the window instead of filtering by search terms; the relevance pass sorts them later. The window is five years, not two, because a count-only probe (no content read) found 1,452 UX and 233 PM posts in two years against 8,762 and 1,974 in five. Comparisons over time must account for the different window.

Study 02: GitLab merge requests (handbook, marketing site, product) with author and reviewer roles from GitLab's public team data. Behavioral data, in place of Reddit (round 4, dropped).

Round 6 adds the voices of product engineers and technical product managers: Hacker News posts where people name the role, either their own ("as a PM", "I'm a product engineer") or the role as a topic ("product engineering", "technical product manager"). The terms are about who's speaking, not about any pain, so they don't favour a hypothesis or candidate. Same two-year window as rounds 1 and 2. A relevance pass later removes posts that aren't about work problems.

Stage 3 deep dive: a random sample of 6,000 merged GitLab product MRs (Oct 2025 to Oct 2026) with AI-disclosure markers from descriptions and commits, plus the review-time trend over two years.