Where we are · last run 28 Sept 2026

Nosy v0.6

Delivery read from --since 2026-09-21 00:00 main · last main commit ace79b8

17of 21 areas done

3partial

1missing

36things shipped this range (125 commits, 0 open PRs)

Latest summary

Cycle 5: Nosy looks "inside" for real, for the first time. The backend inventory counts 174 of 489 endpoints in Tebli as screenless in 0.8 seconds; canwe uses that inventory, decisions, and rivals to give an evidenced verdict on "can we do that?" and writes every answer to the ledger. The Claude Code plugin is ready (/nosy:peek, /nosy:canwe…).
Biggest risk is substitution, our edge is persistence. A founder can ask Claude Code or Cursor directly. Nosy's answer: counting lives in the script (the same question gives the same count), judgment lives in the agent, memory lives in pm/ (the ledger recalls the previous verdict and what's changed since, when the same topic is asked again).
Rivals expanded: 42 products, 21 steps. Code readers (Unblocked, Dosu, DeepWiki) produce a map but don't say "we can do this, it'll take this long"; spec tools (Spec Kit, Kiro, Taskmaster, BMAD) go from decision to code but don't inventory existing code. Nobody fully covers the "backend inventory" step.
The product's name is Nosy. "Nosy about your product. Never your data." Not code review, product decisions: first backend, git and roadmap, then rivals, finally "we can do that, it'll take this long." Nothing to do with Tebli; the repo is now ~/Desktop/nosy.
The tool caught stale info on its own owner's page in Tebli twice (#438 in cycle 2, K180 step 5 in cycle 3). Since cycle 3, findings like this go into the page on their own from peek and psst output, not by hand (the "Today" section).

6 more

No single product does the full cycle. We compared 27 products against a 19-step cycle (42 products, 21 steps by cycle 5): rival tracking, delivery from git, "what can be done now," PRD, roadmap, decision page. Each group does one part of the cycle well; none reaches into the others.
The closest commercial rival is Productboard Spark. It reads the codebase and writes a sourced PRD, does rival research, and has scheduled jobs. But it doesn't read git for "what actually shipped."
The closest agent-native rivals are open-source plugins. phuryn/pm-skills (26.6k stars, 69 skills), Anthropic's knowledge-work-plugins product-management plugin, automazeio/ccpm (8.4k stars; PRD → epic → task → GitHub Issues). All three write PRDs and plans; none does rival tracking or a decision page. The closest to our cycle is ccpm: the only open-source project that uses git and GitHub Issues as the real source of delivery; it's missing rival tracking and a published page. The plugin that most clearly builds the continuous-rival-tracking chain (scan → monitor → comparison card) is deanpeters/product-manager-skills.
Market movement: Amplitude bought Kraftful, the product shut down; Delibr shut down on 31 Aug 2026. A standalone "AI that writes PRDs" struggles to survive on its own, gets absorbed into bigger platforms.
Competitive-intelligence tools are expensive and noisy. Crayon, Klue, Kompyte, Contify run $12–100k/year; the top complaint on G2 is alert noise. Worth taking: a severity score per announcement, an "is this important" classification, visual diffing (Visualping).
Swarmia and Jellyfish are strong on the delivery-reading side, but have no PRD and no rivals. The "link back to the initiative via the key in the commit" pattern we took from Swarmia turned into code in cycle 1 (collect-status.mjs).

By area

Done

1Product and rival definition (setup)

move-in: product.md, decisions.md, sources.json; verify-setup.mjs checks the paths against the repo

8 rivals have it
Aha! AI (+ Aha! Develop, Aha! Roadmaps AI) · Contify · Crayon · Jellyfish · Klue · Productboard Spark · Similarweb · Swarmia
Done

2Scanning rival sites and announcements

neighbors: sub-agents use WebSearch/WebFetch

8 rivals have it
Contify · Crayon · Product Manager Skills — deanpeters/Product-Manager-Skills · Jellyfish · Klue · Productboard Spark · Similarweb · Swarmia
Done

3Continuous rival monitoring, change alerts

watch-rivals.mjs: public landing/changelog/pricing pages vs last snapshot, changed rivals quoted; neighbors re-researches only those; nosy weekly includes it

6 rivals have it
Contify · Crayon · Product Manager Skills — deanpeters/Product-Manager-Skills · Klue · Similarweb · Visualping
Done

4Building the feature matrix on its own

build-matrix.mjs, from the rival files

No rival has it fully
Done

5Demand signal: interviews, support, surveys

collect-signals.mjs: support/chat/call/survey/interview data and GitHub issues matched to matrix rows, requests, decisions and psst items; reads native Intercom/Zendesk/HubSpot/Slack/Gong exports and raw connector responses pulled read-only into pm/signal/ (psst step 2); personal data masked; no Nosy-hosted sync

20 rivals have it
Aha! AI (+ Aha! Develop, Aha! Roadmaps AI) · Anthropic knowledge-work-plugins — product-management · BuildBetter · Craft.io · Product Manager Skills — deanpeters/Product-Manager-Skills · Product Manager Skills — Digidai/product-manager-skills · Dovetail · Enterpret · GitHub Spec Kit — github/spec-kit · Glean Agents · Jira Product Discovery + Atlassian Rovo · Klue · Linear (Linear Agent / Product Intelligence / Code Intelligence) · Pendo · PM Skills Marketplace — phuryn/pm-skills · PM-Skills — product-on-purpose/pm-skills · Productboard Spark · Sprig · Swarmia · Unwrap.ai
Missing

6Demand testing: search volume, landing pages, ads

outbound demand-testing plan written by hand; no tool

1 rival have it
Similarweb
Done

7Reading actual delivery from Git/GitHub (commits, PRs)

peek: git log + open PR commits (full title) + gh

11 rivals have it
Aha! AI (+ Aha! Develop, Aha! Roadmaps AI) · Asking the coding agent directly (Claude Code, Cursor) · CCPM (Claude Code PM) — automazeio/ccpm · BMAD Method — bmad-code-org/BMAD-METHOD · CodeRabbit · Dosu · Jellyfish · Linear (Linear Agent / Product Intelligence / Code Intelligence) · Pendo · Swarmia · Unblocked
Done

8Issue tracker integration (GitHub Issues, Linear, Jira)

GitHub Issues + preread.mjs (with pack patterns); no Linear/Jira

20 rivals have it
Aha! AI (+ Aha! Develop, Aha! Roadmaps AI) · Anthropic knowledge-work-plugins — product-management · Asking the coding agent directly (Claude Code, Cursor) · CCPM (Claude Code PM) — automazeio/ccpm · BMAD Method — bmad-code-org/BMAD-METHOD · BuildBetter · ChatPRD · CodeRabbit · Craft.io · Enterpret · GitHub Spec Kit — github/spec-kit · Glean Agents · Jellyfish · Jira Product Discovery + Atlassian Rovo · Kiro · Linear (Linear Agent / Product Intelligence / Code Intelligence) · Productboard Spark · Sprig · Swarmia · Unblocked
Done

9"What can be done now" inference (delivery ↔ request/gap mapping)

psst.mjs: 6 signals (inventory included), separates the ones already in open PRs; owner's filter is learn.mjs

4 rivals have it
CCPM (Claude Code PM) — automazeio/ccpm · BMAD Method — bmad-code-org/BMAD-METHOD · GitHub Spec Kit — github/spec-kit · Unwrap.ai
Done

10Writing PRDs / specs

gather-evidence.mjs (rarity-weighted) + spill command

17 rivals have it
Aha! AI (+ Aha! Develop, Aha! Roadmaps AI) · Anthropic knowledge-work-plugins — product-management · CCPM (Claude Code PM) — automazeio/ccpm · BMAD Method — bmad-code-org/BMAD-METHOD · ChatPRD · Craft.io · Product Manager Skills — deanpeters/Product-Manager-Skills · Product Manager Skills — Digidai/product-manager-skills · Fleecevest · GitHub Spec Kit — github/spec-kit · Glean Agents · Jira Product Discovery + Atlassian Rovo · Kiro · PM Skills Marketplace — phuryn/pm-skills · PM-Skills — product-on-purpose/pm-skills · Productboard Spark · Taskmaster AI — eyaltoledano/claude-task-master
Done

11Roadmap suggestions, prioritization

build-waves.mjs: open-formula score (Value × Confidence / Effort), waves, pending-decision and not-doing kept separate; size from history via measure-size.mjs

14 rivals have it
Aha! AI (+ Aha! Develop, Aha! Roadmaps AI) · Anthropic knowledge-work-plugins — product-management · ChatPRD · Product Manager Skills — deanpeters/Product-Manager-Skills · Product Manager Skills — Digidai/product-manager-skills · Glean Agents · Jellyfish · Jira Product Discovery + Atlassian Rovo · Pendo · PM Skills Marketplace — phuryn/pm-skills · PM-Skills — product-on-purpose/pm-skills · Productboard Spark · Swarmia · Unwrap.ai
Partial

12Task distribution, assignment to people

measure-size.mjs: owner suggestion from a field's commit share; no assignment (deliberate)

8 rivals have it
Aha! AI (+ Aha! Develop, Aha! Roadmaps AI) · CCPM (Claude Code PM) — automazeio/ccpm · BMAD Method — bmad-code-org/BMAD-METHOD · Craft.io · Jira Product Discovery + Atlassian Rovo · Kiro · Linear (Linear Agent / Product Intelligence / Code Intelligence) · Notion Custom Agents
Done

13Decision log and memory of rules

decisions.md + Claude's memory + canwe ledger + learn.mjs (owner's persistent feedback)

14 rivals have it
Asking the coding agent directly (Claude Code, Cursor) · CCPM (Claude Code PM) — automazeio/ccpm · BMAD Method — bmad-code-org/BMAD-METHOD · BuildBetter · CodeRabbit · Contify · Dosu · GitHub Spec Kit — github/spec-kit · Kiro · Pendo · PM-Skills — product-on-purpose/pm-skills · Productboard Spark · Swarmia · Unblocked
Done

14Publishing a shareable decision page / report

Artifact decision page (Nosy brand); "Today" section from script output, with a stale-line box

14 rivals have it
Aha! AI (+ Aha! Develop, Aha! Roadmaps AI) · BuildBetter · CodeRabbit · Contify · Crayon · Dovetail · Enterpret · Jellyfish · Jira Product Discovery + Atlassian Rovo · Klue · Notion Custom Agents · Sprig · Swarmia · Unwrap.ai
Done

15Citing sources, "unverified" marking

the (not verified) rule; audit-prd.mjs flags unsourced claims

11 rivals have it
BMAD Method — bmad-code-org/BMAD-METHOD · Caveman · Contify · Dosu · Enterpret · GitHub Spec Kit — github/spec-kit · Glean Agents · Linear (Linear Agent / Product Intelligence / Code Intelligence) · Productboard Spark · Unblocked · Unwrap.ai
Done

16Human approval before writing externally

writing outward needs the owner's approval; privacy-scan.mjs scans for secrets and personal data before publishing

9 rivals have it
Aha! AI (+ Aha! Develop, Aha! Roadmaps AI) · Asking the coding agent directly (Claude Code, Cursor) · Caveman · CodeRabbit · Contify · Fleecevest · Kiro · Pendo · Productboard Spark
Done

17Works alongside the code (inside a CLI / coding agent)

Claude Code plugin: /nosy:<command>, sub-agents, SessionStart summary; only the project skill in the cloud

25 rivals have it
Anthropic knowledge-work-plugins — product-management · Asking the coding agent directly (Claude Code, Cursor) · CCPM (Claude Code PM) — automazeio/ccpm · BMAD Method — bmad-code-org/BMAD-METHOD · BuildBetter · Caveman · ChatPRD · CodeRabbit · Product Manager Skills — deanpeters/Product-Manager-Skills · DeepWiki · Product Manager Skills — Digidai/product-manager-skills · Dosu · Fleecevest · GitHub Spec Kit — github/spec-kit · Jira Product Discovery + Atlassian Rovo · Kiro · Linear (Linear Agent / Product Intelligence / Code Intelligence) · Pendo · PM Skills Marketplace — phuryn/pm-skills · Ponytail · PM-Skills — product-on-purpose/pm-skills · Sprig · Swarmia · Taskmaster AI — eyaltoledano/claude-task-master · Unblocked
Partial

18Open source / self-hosted

open-core decision (28 Sep, Bora); release prep on a separate branch (LICENSE, GitHub Action, MCP); not yet released

12 rivals have it
Anthropic knowledge-work-plugins — product-management · CCPM (Claude Code PM) — automazeio/ccpm · BMAD Method — bmad-code-org/BMAD-METHOD · Caveman · Product Manager Skills — deanpeters/Product-Manager-Skills · Product Manager Skills — Digidai/product-manager-skills · Fleecevest · GitHub Spec Kit — github/spec-kit · PM Skills Marketplace — phuryn/pm-skills · Ponytail · PM-Skills — product-on-purpose/pm-skills · Taskmaster AI — eyaltoledano/claude-task-master
Done

19Scheduled runs (weekly report etc.)

desktop-app scheduled task nosy-weekly-stakeout, Mondays 09:00: freshness → nosy weekly (incl. rival watch) → neighbors for changed rivals → diff → log + commit of its own pm/ files; docs/INSTALL.md d0

16 rivals have it
Aha! AI (+ Aha! Develop, Aha! Roadmaps AI) · Asking the coding agent directly (Claude Code, Cursor) · Caddie AI (caddieagent.com) · CodeRabbit · Contify · Crayon · Enterpret · Glean Agents · Jellyfish · Klue · Linear (Linear Agent / Product Intelligence / Code Intelligence) · Notion Custom Agents · Productboard Spark · Swarmia · Unwrap.ai · Visualping
Done

20Backend inventory: endpoint/table ↔ screen mapping, "ready but screenless"

inventory.mjs: 489 endpoints in Tebli, 174 screenless, 0.8s; ~22% false positives in manual check

No rival has it fully
Partial

21A code-grounded, dimensioned answer to "can we do this?" (feasibility)

canwe.mjs + ledger.mjs + measure-size.mjs (size from history; confidence from coverage and field consistency); the agent verifies matches

No rival has it fully