✓ You're ~80% there already

The Idea → Project Engine

You've already built 80% of this. Here's the one missing piece — and the smallest way to add it without rebuilding anything you have.

The saved reel

1What the saved post actually describes

Grounding

2What you already have — OTTO

Your system already runs option (B) in substance. Every piece below is live today.

✓
PM brain (OTTO)
Thin, restartable orchestrator. Writes briefs, gates results, sequences work — never does the work itself.
✓
Never-forget board (QB_BOARD)
Read first, written last each session: shipped, in-flight, queued, blocked, open decisions. The dashboard mirrors it.
✓
Disposable workers
Born → one job → die. Tiered Mechanical / Standard / Deep (best-of-N + judge + skeptic).
✓
3 subscription accounts
1 laptop anchor + 2 heavy lanes on a Mac Mini via a gateway. Jobs route to whatever has headroom.
✓
Second-brain vault
Obsidian, numbered folders, per-project plan + memory files, reusable role files (builder/verifier/researcher).
✓
Dashboard (otto.bbase.ai)
Per-project chat, a "needs-you" strip, usage meters, live ticker, two-way inbox with an always-on catcher.
✓Idea capture already exists. Your Instagram saves are auto-pulled into the vault (the ig-saves workflow) — so "capture ideas from content I save" is already a working feed.
The gap

3The one missing piece

There is no clean lane that turns a raw idea into a scoped, scaffolded, running project.

Today ideas land in chat or the dashboard inbox and you promote them by hand — deciding if they're worth it, writing the brief, making the project, kicking off the first worker. Your own board lists "idea → task → project promotion" as still pending. That single hand-off is the whole job below.

The recommendation

4The recommended pipeline

An idea flows top-to-bottom through these stages. Only one stop needs you — it's outlined in blue.

1
Capture automatic
Any idea — from chat, the dashboard inbox, a voice note, or your IG-saves — lands in one shared Idea Queue. Who: existing inbox catcher + ig-saves feed.
2
Triage & score automatic
A worker rates each idea (value, effort, fit, urgency), checks it against your ~25 live projects, and quietly parks or merges the weak ones. Only real candidates move on. Who: disposable triage worker.
3
Scope — auto-draft a mini-brief automatic
A worker writes a one-page plain-English brief: the goal, what "done" looks like, the first 3 tasks, which effort tier, which account lane. This is the thing you'll look at. Who: disposable scoping worker.
★
Your one-tap approve the only stop that needs you
The mini-brief appears on your "needs-you" strip. Approve, tweak a line, or reject — one tap. Nothing scaffolds or spends until you do. Who: Loren.
4
Auto-scaffold automatic
On approval: a plan file + project memory file are created, a QB_BOARD row and a dashboard row appear, and the project is real. Who: scaffold routine on OTTO's plumbing.
5
First worker fires automatic
The first task dispatches to whichever account has headroom and starts producing. Who: existing worker dispatch + gateway.
6
Tracked to done automatic
Progress lives on QB_BOARD and the dashboard like every other project — checkpoints return to your needs-you strip only when something is irreversible or outward-facing. Who: existing board + dashboard.
The delta

5Reuses what you have vs net-new to build

♻ Reused from OTTO, as-is
✓PM brain — sequences & gates the new lane
✓QB_BOARD — the new project row lives here
✓Disposable workers + tiers — run triage, scope, build
✓3 accounts + gateway routing
✓Vault, plan files, memory files, role files
✓Dashboard + needs-you strip (hosts the approve tap)
✓Inbox catcher + ig-saves (feed the queue)
+ Net-new (small — this is the whole build)
Idea Queue + idea format small
One shared list every source writes into, in a consistent shape.
Triage & score step small
Rates ideas and de-dupes them against your 25 projects.
Scoping → mini-brief medium
Drafts the one-page brief you actually approve.
One-tap approve control small
A button on the strip you already have.
Auto-scaffold routine medium
Turns an approved brief into files + rows + first worker.
The fork

6The decision you actually face

How should the agents be organized? Three honest options — one wins for your situation (one founder, ~25 projects, 3 subscription accounts).

Standing directors
4 always-on agents (eng, marketing, research, growth). Faithful to the reel — but each one burns subscription headroom around the clock and adds a coordination layer you don't need yet.
A team per project
A fresh standing team for each idea. With ~25 projects that's a lot of idle teams competing for 3 accounts.
★ Recommended
Hybrid
Keep OTTO's proven brain + disposable workers. Treat "directors" as on-demand role hats (builder/verifier/researcher), not standing agents. Add only the idea→project lane.
Why hybrid, for you specifically: 3 subscription accounts can't fund four always-on directors and the workers doing real work — standing agents would starve the actual jobs. Your owner-session + worker model already delivers ~80% of what "directors" promise. Standing directors only pay off at a scale you're not at. So: graft the reel's PM-on-top idea (you have it) without paying for its standing-director overhead.
1. Standing directors, project-teams, or hybrid?
Hybrid — workers under OTTO's brain, no standing directors.
Why: your headroom is the real limit; standing agents would compete with real work.
2. Should good ideas auto-promote, or always wait for you?
Always require your one tap — never auto-promote.
Why: keeps a single human gate, costs you one tap, and stops runaway projects.
3. One idea queue, or capture per-source?
One shared Idea Queue all sources feed.
Why: ideas scattered across chat, inbox, and IG is how good ones get lost.
4. How aggressive should triage be?
Aggressive — auto-park/merge weak ideas, surface only real candidates.
Why: your attention is scarce, agent effort is cheap — spend the cheap one.
Your side

7How it feels for you

  1. You have an idea — say it in chat, drop it in the inbox, or just save the reel. Done on your end.
  2. Later, a clean one-page brief is waiting on your needs-you strip (the weak ideas never reached you).
  3. You read it in 20 seconds and tap approve (or tweak a line, or reject).
  4. The project is now scoped, on the board, and already working — with no further steps from you.
Sequencing

8Build path

Phase 1 · Proof
One idea, end to end
A single idea in the queue → triage + scope draft a mini-brief → it hits your strip → you tap approve → one worker fires. Manual edges are fine.
Why first: proves the approve-gate + auto-scaffold loop works before you generalize anything.
Phase 2 · Full pipeline
All sources, all 25 projects
Wire chat, inbox, and IG-saves into the shared queue; score & de-dupe against every live project; tier + lane routing; tracked-to-done on the board.
Why next: only worth generalizing intake once the core loop is proven on one idea.
Phase 3 · Product
An operator-in-a-box
Package it as a Brightbase product: per-client vault/board/queue, a reusable scaffold template, capture from anywhere. Generalize the queue + scaffold; keep your routing hardcoded.
Why last: only sell a workflow you've proven on yourself first.
Skeptic pass

9Honest risks — where this could over-build

★ The single smallest first step
Add a shared Idea Queue, and have one worker draft a mini-brief for one idea and drop it on your needs-you strip for a one-tap approve. Everything else can wait until that one loop feels good.