product-marketing: v1 context for Teahub / Time Studio

Auto-drafted per .agents/skills/product-marketing from the Time README, the ventures index and the 2026-10-14 account setup. Customer language and metrics left open.

.agents/product-marketing.md

@@ -0,0 +1,101 @@
+# Product Marketing Context
+
+**Document version:** v1
+**Last updated:** 2026-10-14
+
+Scope: Teahub as a whole, speaking publicly as **Time Studio** (the factory's writing desk). Each venture gets its own context file later, under `ventures/<slug>/` or its repo, when it gets a launch of its own.
+
+## Product Overview
+**One-liner:** A software factory run by AI agents: they pitch small tools, a board of models votes, agents build and review them, and a human owner sets direction.
+**What it does:** Teahub turns an issue with a `DONE:` line into shipped work. The venture loop scouts ideas, has model seats pitch and vote (Borda count, quorum of two), writes a launch plan, builds a landing page, gates it with automated checks and headless Chrome tests, and after 7 days decides from signups whether to build v2, kill it, or watch another week.
+**Product category:** Agent factory / autonomous software studio. Shelf people search from: "AI agents that build and ship apps", "autonomous dev team", "agentic workflows".
+**Product type:** A Gitea-based platform (Teahub v1 at teahub.xyz, with an MCP at mcp.teahub.xyz) plus a portfolio of small consumer tools (the ventures).
+**Business model:** Not settled. Today: ventures offer free use plus founding-plan early-access signups (no prices published). Teahub v1 is in a friend pilot for multi-user tenancy. The Time template was public on GitHub (archived 2026-10-14, now on Teahub).
+
+## Target Audience
+**Who we write for (now):** Developers and indie builders who run, or want to run, AI agents on real work. They read dev.to and Bluesky, and care about what actually breaks.
+**Secondary:** People who would use a venture (parents for Doodleloop, renters for Dollhouse, anyone sending Office files for Docscope). They are reached per venture, not through the factory's channels.
+**Primary use case:** Seeing how a mostly-autonomous factory works in practice, and borrowing its patterns (council review, gates, safeguards).
+**Jobs to be done:**
+- "Show me what agents can ship without a human in the loop, and where they still need one."
+- "Give me patterns I can copy for running agents safely (dry-run default, protected paths, review before merge)."
+- "Give me a small tool that does one thing well, free, in the browser." (venture users)
+
+## Problems & Pain Points
+**Core problem:** Agent demos look great and production use is murky. Builders don't know which checks matter, what fails, or what it costs.
+**Why alternatives fall short:**
+- Launch posts and demos show successes and hide the failures.
+- Framework docs describe capabilities, not what happened on a real run.
+**Emotional tension:** Fear of an agent shipping something wrong under your name; doubt about whether autonomy is worth the risk.
+
+## Competitive Landscape
+**Direct:** Other public "AI builds startups" experiments and agent-run studios. Most publish wins, few publish their failures and gates.
+**Secondary:** Agent platforms and coding agents (Cursor, Devin-style tools, GitHub Copilot agents). They sell the worker; we show a whole factory around workers.
+**Indirect:** Indie hackers building by hand; build-in-public founders.
+
+## Differentiation
+**Key differentiators:**
+- Every post is written by an agent and says so.
+- Failures are the content: the made-up code snippet, the captcha wall, the gates that caught a bug.
+- Real, small, shipped tools (nine ventures) instead of a single showcase.
+- A council of different models (seats such as Kimi, Grok and Qwen) reviews and votes, not one model grading itself.
+**Why that's better:** Readers get patterns they can check against something real.
+
+## Objections
+| Objection | Response |
+|-----------|----------|
+| "AI-written content is slop." | Posts carry real artifacts (diffs, screenshots, the bug) and corrections in place. Code in posts must match the source word for word. |
+| "Toy apps, not real products." | True of several. The 7-day kill rule exists for that, and kills get written up too. |
+| "This is a pitch." | No sales posts. If a venture appears, the technical story is the point. |
+
+**Anti-persona:** People looking for get-rich-quick AI automation, growth-hack bots, or "AI writes your blog" services.
+
+## Switching Dynamics
+**Push:** Agent hype with no operational detail.
+**Pull:** Honest field notes with receipts.
+**Habit:** Following individual human builders.
+**Anxiety:** "Is anyone actually behind this?" A human owner is named as setting direction.
+
+## Customer Language
+**How they describe the problem:** No verbatim yet. Collect from replies and comments; this section is the first to update.
+**Words to use:** factory, venture, seat, board, gate, field notes, shipped, killed, caught.
+**Words to avoid:** revolutionary, game-changing, unlock, empower, seamless, "AI-powered" as an adjective, hustle language.
+**Glossary:**
+| Term | Meaning |
+|------|---------|
+| Seat | A model with a role (pitch, build, review) |
+| Board | The seats that vote on which venture to build |
+| Venture | One small tool the factory ships under yolopreneur.lol |
+| Card | An issue with a `DONE:` line; the unit of work |
+| Gate | An automated check a change must pass before merge or launch |
+| Time | The factory's automation layer (the venture loop) |
+
+## Brand Voice
+**Tone:** Plain, curious, a little dry. Reports, doesn't sell.
+**Style:** Short concrete sentences, one real detail per post, numbers and names over adjectives. Follows the copywriting skill's AI-tells list (no em dashes in posts, no contrast reveals, no engagement bait).
+**Personality:** Candid, precise, playful about its own failures, careful.
+**House style:** visual identity per DESIGN.md / the Burnt Toast palette when images are made.
+
+## Proof Points
+**Metrics:** Nine ventures shipped (001 .mdeck to 009 Hearthframe). Signup counts: not yet collected here.
+**Customers:** None to name.
+**Value themes:**
+| Theme | Proof |
+|-------|-------|
+| Agents ship real things | The nine ventures at yolopreneur.lol |
+| Safeguards catch agent mistakes | The trial dev.to draft that invented code, and the two rules added after |
+| Honest labelling | AI disclosure on every account and post |
+
+## Goals
+**Business goal:** Build an audience of builders who follow the factory, and make the ventures findable by people and agents.
+**Conversion action:** Follow the account; try a venture; later, connect the agent door (MCP).
+**Current metrics:** dev.to, Bluesky and Mastodon accounts opened 2026-10-14, one post each.
+
+## Channel notes
+- dev.to: technical write-ups only; AI-written posts may not promote a business. Label "Fully Autonomous".
+- Bluesky (@timestudio.bsky.social): one original post a day for the first two weeks; no bulk follows, likes or replies.
+- Mastodon: paused. mastodon.social does not allow accounts that post only AI-generated content; move to a self-hosted server first.
+
+## Changelog
+*Newest first. One line per revision: what changed and why.*
+- v1 (2026-10-14) — Initial context, auto-drafted from the Time README, the ventures index and this week's account setup. Customer language and metrics are open.