Core Thesis
LLM Wiki for Your Team
"A persistent updating knowledge base for your team that is well-structured."
BackboneCommitted
What gets tracked
- Action items — auto-tracked in meeting notes and team/company task list
- Ideas — added to meeting notes and company idea base
- Conflicts & gripes — long-standing Wikipedia-style relationship pages
- Pattern detection (e.g. "didn't respond for 3 days" → flagged as instance of existing gripe)
- Vision documents — every add updates the big vision doc
- Challenge/issue patterns — documented when recurring
Structural detail (later)
- Wiki-hub structure vs. Noos graph structure?
Workflow Open Question
Input Pipeline
"Really nailing the workflow from wherever you record meeting notes to updating this is a question we need to figure out."
Scope Decision
Existing note apps all fragment
- Gemini meeting notes → Google Doc
- Granola → elsewhere
- Otter → elsewhere
The fork
- Hyper-focus the whole team on nailing one workflow end-to-end might be right thing
- Continue parallel product tracks (see below)
Parallel Track
01Slack Bot
"Essential. Table stakes."
Table StakesCommitted
- References team's LLM wiki
- Surfaces information
- Pulls from both written history and meeting history
Parallel Track
02Group Chat Bot
"A bot you can add to an arbitrary group text that maintains knowledge from it."
ExploringScope Decision
Platform candidates
- iMessage "very important, at least in the US"
- WhatsApp — TOS workaround: advertise as "human assistant service" (assisted by AI) to avoid bot ban
- Signal, etc.
Architecture fork
- Shared bot — a number you add to the group chat
- OpenClaw-style — no bot in group; runs on your end, maintains knowledge from your own history
- Both — there may be a place for each
Cross-privacy scoping
- Bot can share some of your info, not all
- Asks for approval per share
- e.g. "Share your availability on this date to help schedule?" → you click Approve
Open question
- Are there high-signal WhatsApp groups worth retrieving for? (asked Max)
Parallel Track
03Real-time Meeting Outliner
"Custom UX — transcribes, takes structured notes, updates your KB every meeting."
In ProgressScope Decision
- Already being built
- Custom user experience
- Updates LLM wiki with every meeting
Strategic question
- Build our own note-taking app vs. integrate with existing ones people already use?
Parallel Track
04Ambient AI Display Dashboard
"Forget meeting note-taking — just an AI ambient display dashboard that keeps track of things."
Could Be Own Product
- Pops up relevant info during meetings in real time
- Does research, fact-checks, builds stuff live
- Accepts real-time prompts
- "Give me a dashboard of all key branches we could go down" → this page is literally an example
- "Hey AI, go build that node right now" → spawns full product
Strategic question
- How important is real-time in-meeting assistance vs. just tracking commitments?
Parallel Track
05Note Stream (mobile app)
"Slack for myself — collaboration-first, LLM-friendly version of ThoughtStream."
Product Direction
- Evolution of current App Store app (ThoughtStream)
- Capture, tag, filter notes
- Spawn arbitrary agent per note
- Go build something
- Flesh out an idea in the thread
- Doubles as an agent deployment dashboard
Parallel Track
06Better GPT
"ChatGPT, but Claude Code / Codex under the hood, on a private server."
Product Direction
- Coding agent under the hood (Claude Code or Codex)
- Runs on private server
- Does everything a coding agent can that normal GPT can't
- Connectors to normal ChatGPT and Claude
- → gives those a "cloud box" with full agent capabilities
Scope Decisions on the Table
What We Must Cut Or Pick
"A lot of overlapping use cases. Be precise about which one we're talking about."
Blocking
Market shape
If B2B
- Integrate with Google Meet?
- Integrate with Zoom meeting notes? How?
- Build our own meeting note-taker?
If group-chat bot
- WhatsApp, iMessage, or Signal first?
- OpenClaw-style (your own bot, your own history) or shareable number added to groups?
On the product itself
- Custom note-taking app or integrate with existing?
- Real-time ambient display — essential, or nice-to-have?
- Wiki-hub or Noos graph structure?
"It might come down to 80/20s in the actual market where one of these works first. But good to think of the possibilities and figure out which are most essential."
Why Not Pick Just One
Shared Infrastructure & Synergies
"Every one of these is valuable. Every one is synergistic. A lot share infrastructure."
Platform Bet
Common backbone
- LLM Wiki (Track 0) is the shared knowledge layer every surface reads & writes to
- Slack bot queries it
- Group-chat bot queries it
- Meeting outliner writes to it
- Ambient display surfaces from it
- Note Stream reads/writes personal slice
Cross-surface agents
- Better GPT infra = private coding-agent server
- Powers "go build that node" from ambient display
- Powers per-note agent spawns in Note Stream
Privacy / scope model
- Cross-privacy-scoped bot pattern (approval-per-share) reusable across iMessage, WhatsApp, Slack, email
Meta capability
- The dashboard you're reading right now is itself a demo of the ambient-display product
- Prompt: "give me a dashboard of all key branches"
- Output: this page
- Next step imagined: "hey AI, go build that node" → product gets built