COS Pipeline — What Gets Captured

📧 Email

Gmail or Outlook · scanned every 2 hours on weekdays

Fully automatic

Source

Your Gmail or Outlook inbox — every inbound email since the last check. Your inbox only; CC'd emails are picked up automatically.

Pipeline

Pass 1 — Fast sort (Haiku) — every email categorized: Deal / Action / Research / Ignore

Pass 2 — Deep read (Sonnet) — Deal and Action threads only: what's the ask, what do you owe, what do they owe

Drive output

Follow-ups Doc — action items, tagged by owner and deal

Pipeline Doc — deal threads

People Doc — new contacts named in threads

Dashboard

Follow-ups tile — new items sorted by due date and owner

Pipeline tile — deal threads updated

🎙 Calls & Transcripts

Daily at 7:22am + real-time webhook when a transcript appears in Drive

One-time setup

Source

Your recording service — Otter AI, Beside AI, Fireflies, Grain, Read AI, Fathom, or the built-in desktop recorder. Syncs transcript as a text file to your configured Drive folder after each call.

Supported formats

.txt .vtt .srt .md .rtf Google Doc Ignored: .m4a .mp3 .wav .pdf

Pipeline

Pass 1 — Investor memo (Sonnet) — 6-section memo: core argument, what was agreed, open questions, what you'd need to form a view, names & firms

Pass 2 — Action extraction (Opus) — every commitment named: who said they'd do what, by when, who owes whom

Drive output

Pipeline Doc — full 6-section memo appended

Follow-ups Doc — action rows per commitment, owner tagged

People Doc — new contacts from the call

Dashboard

Follow-ups tile — your actions + what the other party owes you

Pipeline tile — deal and LP intelligence updated

People tile — new contacts appear

→

One-time setup: connect your recording service to Drive Otter AI — set up Zapier: New Otter recording → Upload to Google Drive folder using your folder ID. · Beside / Fireflies / Grain / Read AI / Fathom — enable native Drive sync in that service's settings, point at your folder ID. · Manual — paste transcript into a .txt file, save to any configured Drive folder; filename becomes the call title.

🎧 Podcasts & Blogs

RSS feeds checked daily at 6:45am · episodes transcribed and summarized overnight

Automatic after config

Source

Your RSS feed list — configured under personal.content_feeds in firm_context.yaml. Each team member has their own list. New episodes are detected daily.

Pipeline

AssemblyAI transcribes the audio — ~5 min per episode, returns full transcript text (~$0.009/min)

Claude writes 6-section memo (Sonnet) — core argument, consensus, tension, open questions, names & firms

Deal scan — memo checked against your tracked deals; relevant intel tagged as a deal takeaway

Drive output

Per-show Google Doc — full memo prepended; each show has its own doc

Deal-tagged intel — takeaways routed to relevant deal tiles

Dashboard

Market Intelligence tile — new summaries with deal-relevant flags

Morning briefing (7:51am) — new episodes summarized, deal flags called out

🚫 What Does Not Get Captured

Nothing below is ever touched — no access, no indexing, no processing

Out of scope

What's excluded

Why & workaround

PDFs — pitch decks, IC memos, research reports

Not yet supported. Planned feature. Content surfaces through call transcripts where those docs are discussed.

Drive files outside your configured folders

Drive API only queries the specific folder IDs in your transcript_sources. Add the folder to your config to include it.

Slack / Teams messages

No integration. Workaround: copy key threads into a .txt file and save to a configured Drive folder.

Calendar event content & meeting notes

Calendar is read for scheduling context only — not for content. Use a recording service that auto-joins and transcribes.

Other people's inboxes

Pipeline reads your inbox only. CC'd emails reach you automatically.

Local Desktop files (not in a configured folder)

Add a local_folder source to transcript_sources pointing at that path.