A vision for Ideaflow · Draft 03 · Closer to the recording
Give people the space to flourish, and to be operating and living from a coherent state of mind.
Creating a space with your personal knowledge system that you would proudly give a young, bright child who’s coming into the world.[1]
A nice place to capture your insights, and an organized, spacious location for all the precious things you want to learn. Your own Zen garden. The captain’s seat in your creative studio—from which you can communicate with humans and AI agents, and take action.
What happens when many people are operating from these mind studios? We can enable collective intelligence: individual note-taking, connected to a collective knowledge base that, subject to privacy scope, different people can add things to.
Business targets and practical matters aside for a moment: this is the destination.
The relationship that the internet currently has with our attention is not something I’m entirely happy with.
If I had a child, what technology would I actually be proud to show this person? I like the metaphor of thinking of how you’d care for a small child, because it drops me into this meditative state. Here’s a nice place to capture those good insights that you have. Here’s an organized, spacious location for all of the precious things that you want to learn. Inside this simple baseline insight, I see an entire product direction.[1]
You come across something really precious or beautiful, and you can put it on a shelf—a digital shelf for yourself. If you have a skill that you want to learn, you can put the video for learning that skill onto this digital shelf space. We expect to give people a room in this world. We expect to give them a phone and some capabilities. This is the natural evolution of that impulse: to give people the space to flourish.[1]
How can you work without leaving your own Zen garden, your own sanctum, your own state of coherence of mind—without becoming totally carried off in the chop? You should be able to capture your thoughts and even send them out without going into a messy, chaotic environment that is an attentional minefield. You shouldn’t have to wade through the tabloids of the internet to start to interact with things. A world where we’re living in coherence instead of panic is both healthier and wiser.[2]
How do you command the world from your notebook and your mind studio? How do you sit in the control room, or the captain’s seat, in your amazing creative studio? A piece of this is being able to communicate with humans from this state, and also other types of minds, which include AI agents. I see AI agents and humans as the same type of entity, effectively. You can have distinctions between them, but a lot of the infrastructure is shared: they’re both minds, and they both can take action.[2]
That’s part one: give this to everybody in the world, so they carry this amazing creative studio with them. Part two is what happens when many people are operating from these mind studios. There are some very nice things we can do to enable collective intelligence. A knowledge system can track who people are, where you met them, what they’re needing, and what they could offer to your network. That bridges the gap between a personal knowledge system and a collective one.[3]
The destination is an agent-friendly, open social network. That aspect of our company, I think, needs to be fully open source. If I had to choose a single thing, the B2B aspect is secondary to building a collective brain for humanity. That has two parts: the individual note-taking thing, and this collective knowledge base that, subject to privacy scope, different people can add things to.[4]
I think of myself as a gardener for the internet, to help it grow in the healthiest way possible. I don’t trust that the world will just end up in the state that I’m wanting, even if we let a year and a lot of products go by. I think it’ll get better. But what I can bring to the table is seeing the target that we really want to end up with, with a little bit more clarity.[5]
In the recording, the personal knowledge system itself has two parts: the stream where you capture thoughts, and a structured knowledge graph of the facts and relationships that come out of them.
A place for thoughts as they arrive. You can begin with what is on your mind.
A beautiful discovery. A video for a skill you want to learn. A place to put it and come back to it.
Who people are. What they’re needing. What they could offer. Knowledge that helps people take action together.
“What technology am I actually
proud to show this person?”
Exact wording from the automatic transcript · Passage beginning 00:34
I think we are shifting into a world where the most beautiful vision wins. My contribution is three open things.[6]
A method for taking notes, like GTD or a modern Zettelkasten: quick capture, tags, per-line filtering, links. A pattern you can use in an app, a chat, a text file, or a paper notebook. Readable by you, and by your personal agent, in whatever surface you choose.
An open, agent-friendly alternative to LinkedIn.
A WhatsApp alternative with a back channel for agents. iOS App Store · chat.ideaflow.app
The Thoughtstream Protocol: How to Thoughtstream on Any Surface →
A standalone working article in Jacob’s words: the simple practice, examples, and optional advanced conventions.
Note taking is a behavior: getting thoughts out of your brain seamlessly. Here is a protocol for Thoughtstreaming, a way of augmenting your mind with the computer, and what follows once your agent can read it.
Any thought you have, you should be able to capture as if you were writing on a piece of scratch paper: one button press or fewer, for instance dictation. The capture surface can’t limit your messiness. Text gets written as scratch. Audio keeps the original recording, in case the transcription screws up.[6]
The ingredients are seamless quick capture, tags, and filtering by tag on a per-line basis, where you can jump back to any line to see its context. That is what is sucky about Markdown files and Obsidian: you can’t filter line by line. Linking and relations matter too. This is a method, like GTD, a modern Zettelkasten, and it’s worth teaching people how to do it. You could write a tutorial on it, even a book. You can do it with spoken thoughts: dictate the tags along with the note. You can even do it inside chats. You could do it on WhatsApp if you wanted to, and OpenChat supports it. Now that AI is here, tags are in some sense optional, but not entirely, and they aren’t going to be replaced. Tags are an intermediate language. Sometimes AI can tag things for you, but you still need the structure, you want really fast lookup, and you want to define your own structure. There are also some common tag structures that make sense, like what’s relevant and what’s reference.[6]
This is a method for taking notes, independent of any implementation and any app. You can approximate it with a pile of Markdown files, but it gets ugly: you need a new file for each day, you can hit Command-A and accidentally delete everything, and you’re borrowing the ergonomics of the physical world, paper in files. The mind doesn’t work like that. The mind is a Thoughtstream. The Buddhists even have a word for it, the mindstream, almost as a base property of the universe. Our digital tools have the potential to reflect that, so they should.[6]
You should be able to store and browse your Thoughtstream whether you’re using Ideaflow, Mem, Logseq, or whatever comes next. And everyone is about to have a personal agent. It’s obvious that this is the future. These agents need to be able to read your stream of notes, and an entirely new class of software becomes possible once they can. The same goes for longer voice recordings: that tool needs to interoperate too. Any company trying to defend context is fighting a losing battle. Your agent should have all the context of your life. From OpenClaw to Hermes to the assistants every big platform is shipping, everyone is going to have one of these. How the future interoperates is the question.[6]
LinkedIn and Facebook are effectively CRMs: how you keep track of the people you meet, a collective CRM. Because they’re defending the moats of their networks, they aren’t making themselves agent-friendly. An agent-friendly equivalent is going to win, or the main players will be pushed to become agent-friendly. And if they do, they lose the moat anyway, because federated systems where you choose your own algorithm and your own way to browse can be built easily off open APIs. With computer use, those APIs are already less defensible. The scraper versus anti-scraper war has been won definitively by the scrapers. People can align with that reality or not.[6]
Recommender systems that nudge you into greater coherence are huge alpha. The folks building engagement algorithms instead of empowerment algorithms are short-sighted: they’re trying to keep people on the screen, but people wake up to the fact that being sucked onto the screen may not be to their highest benefit. If we build a more wholesome alternative, a small community can expand outwards. And as agents control a whole computer, they can scroll your feed for you. Everything is about to have an API, as Peter Steinberger put it. Everyone’s feed is about to be filtered through their personal agent, or at least we have the possibility for that.[6]
There needs to be an open, agent-native chat app that isn’t fighting the bots but rolling with what they’re trying to create. Every chat should have a back channel where my agents can talk to your agents, as it were. In that channel we can share, with total abandon and without spamming anybody: events we want our friends to know about, asks and offers, a Burning Man ticket for sale, a room opening in a group house, a search for a housemate. Right now group chats and direct invites are one of the primary ways I hear about what’s coming up. We could 10x that. An agent can send a workshop invite to every potentially relevant group chat, because it isn’t distracting any humans. Your agent makes sense of it and tells you only when it’s relevant: “Hey, you should know about this event.” When there’s a coincidence of needs, someone selling a ticket and someone looking for one, it notifies both of you. You click approve, and your agent drafts the message and facilitates the coordination.[6]
Every time you go to a conference, there are tons of intellectual soulmates who don’t meet each other. A system can suggest all the people who should meet, based on overlapping coincidences of needs, mediated by their personal agents, which can share asks and offers, or goals and offers, at different levels of granularity. That is the scale of collaboration personal agents unlock: humanity-wide coordination, and also just your local friend group.[6]
Everyone is about to have a personal agent.
How the future interoperates is the question.
A more enlightened, more conscious internet isn’t one app. It’s a set of potential interlocking components, agent-friendly, that together could cover the basic utilities we use every day, so you barely have to leave the ecosystem.[7]
These are candidates, not a roadmap. Each borrows a form people already know, so nothing needs explaining. What’s new is underneath: one sign-in, one graph, and a back channel your agent can use. Here is each component, the project that realizes it, and where it actually stands as of October 2026.
LinkedIn is a CRM for your professional life that won’t let your agent in. Unlinked is the open version: bring your LinkedIn export, get a public profile, and keep your network as data you and your agent can actually use. Messages run on OpenChat underneath, so the two are already one network.[7]
Liveunlinked.ai
Partiful shows the host a list of names. An event page should know who is in the room. Ideaflow Events is a Partiful-shaped event page where you and your agent are both guests: RSVP in one tap, opt in to matching, and before the party your agent tells you who to talk to and why. Superconnector was our earlier run at matching people at events. This is where that work goes.[7]
PlannedPlan written, not built. Superconnector is not currently running.
Events live in Partiful, Luma, group chats, Instagram stories, and email. A universal events navigator pulls them into one view: what’s on, who you know who’s going, and which ones you’d regret missing. It overlaps with Superconnector and Superconnector Events, which did the who-should-meet-whom part, and with Ideaflow Events, which is the page itself. The navigator is the guest’s side: every event, not just ours.[7]
IdeaNot built.
Bloomscroll is the anti-doomscroll: a scroll of your own saved material, the notes you tagged to remember, and the sources you chose, with spaced repetition built in. Keep what matters and it enters your notes. The rest passes. A feed should be judged by the state you’re in when you put the phone down.[7]
Private previewRunning for us. Not public yet.
GitHub Issues works because problems are visible, structured, and anyone can pick one up. World Issue Tracker does that for problems outside software: make them visible, organize the work, and keep the record public.[7]
Most of what a network is for is a coincidence of needs: someone selling a ticket, someone looking for one. In OpenChat you can publish an ask or an offer as an anonymous intent. It matches quietly, shows only the relevant terms, and reveals nothing until both sides approve. The next step is making asks part of your Unlinked profile, with each field carrying its own audience: everyone, your connections, or only you.[7]
Live in OpenChatProfile layers on Unlinked are designed, not built.
OpenChat is a WhatsApp-shaped chat where agents are participants, not intruders. Use whatever client you want: Unlinked’s messages are their own interface on the same network, and any agent can connect over MCP. The back channel is where my agent talks to yours, finds the overlap, and only tells us when it matters.[7]
Capture a thought with one press or fewer, tag it, find the line again, and send it to a person, a team, or a collection without leaving the note. That is Ideaflow Notes and Thoughtstream Vision. The method behind them is the Thoughtstream Protocol, and it works on any surface.[7]
Liveideaflow.app · vision.ideaflow.app (early) · The protocol
Write “add Sam to the newsletter” and Sam is on the list. Note that you’re about to start a deep-work block, and the computer sets itself up. The note stays a note. Your agent reads it and acts, and asks before anything touches the outside world. We have built the first slice, where Vision reads a note and proposes what to update, and it is behind a flag.[7]
In progressNot shipped.
None of this requires leaving the tools you already have. Every piece here exposes itself to agents, so your agent can glue it to Gmail, your calendar, or anything else with an API or a screen. The apps don’t need integrations with each other. Your agent is the integration.[7]
PatternThe agent surfaces exist today. The automations are yours to write.
You shouldn’t have to give ChatGPT, Claude, and every new assistant separate access to your Gmail. Ideaflow ID is one sign-in across all of these apps, and the shared connector at id.ideaflow.app is one connection that gives your agent OpenChat, Unlinked, and Vision, with permissions you choose. We built the version for our own apps. The general version, an omniconnector for everything, is something the industry should build once, and people are already discussing it.[7]
Each of these apps is a view of the same graph: people, ideas, issues, events, with per-account privacy. That graph is Noos. World Issue Tracker already stores there, OpenChat and the private notes you keep on people share it, and the rest is moving. One backend means a note written in one app shows up in the other, and your agent asks one place.[7]
DirectionPartly in place. Public entry point base.ideaflow.app is coming soon.
This is a first-person editorial draft, not a verbatim transcript. It keeps Jacob’s vocabulary, metaphors, sequence of ideas, and much of the original phrasing. Repetitions, fillers, false starts, and incomplete sentences have been removed; pronouns and sentence order are occasionally adjusted. Headings and transitions were assembled from the discussion. Quotation marks are reserved for exact transcript excerpts.
The supplied recording is #ExplainingIdeaflow Lunch. Otter’s speaker labels are inconsistent; attribution follows Jacob’s identification of this vision as his own and the conversational context, not verified speaker diarization. Timestamps locate passages, not precise word boundaries.
“an organized, spacious location you can hang all of the precious things that you want to learn”
The first two principles closely follow this passage; the manifesto heading and opening condense its wording.
“How do you be in the sort of control room or the captain's seat in your amazing creative studio”
Principles three and four retain the metaphors and human/agent framing. “You shouldn’t have to” restates the criticism of having to wade through incoming demands.
“part two is what is happening when many people are operating from these mind studios.”
The two-part overview and fifth principle combine adjoining passages. The investor story is summarized, not independently verified.
“building a collective brain for humanity”
The sixth principle joins two passages. Open source is scoped to the network aspect; privacy scope is retained. These are aspirations, not claims about available product features.
“I don't trust that the world will just end up in the state that I'm wanting”
The final principle brings together the opening gardener metaphor and the closing statement of intent. Named competitor criticism is omitted.
“any thought you have to be able to capture and write totally seamlessly, as if you're writing on a piece of scratch paper”
The addendum comes from two separate dictated notes (one an incomplete Otter export, the second on tagging and filtering) and an Instagram story, not from the September 17 recording. It keeps the sequence and phrasing of the note; product names the transcript rendered unclearly are omitted rather than guessed. The three contributions and their links follow the story. Product links describe entry points, not claims that every capability described is available.
“a set of interlocking, agent-friendly pieces”
Section 05 comes from a separate dictated note, lightly cleaned, not from the September 17 recording. It keeps the list and phrasing of the note, with later additions from Jacob on October 11 (the framing as potential components, the events navigator, and the reputation aside); the project names and status lines were checked against the live products and repositories on October 11, 2026. “Live” means there is a public entry point you can use today. “Planned,” “In progress,” and “Direction” mean exactly that. Product links describe entry points, not claims that every capability described is available.
The original recording and extracted transcript are retained privately. This public working draft includes editorial source notes and passage timestamps.
Read draft 01 · Read draft 02 · Explore the learning dashboard mockup