Intelligence for Cyborgs.
You read, listen, and talk to people all day, and the parts worth keeping scatter: a line from an article, a point from a meeting, a thread you want to remember. Soulvant is the one place where all of it is gathered, organized, and connected for you, and kept current by an AI assistant that does the upkeep.
You bring what matters; the assistant reads each piece, writes it into clear pages, links those pages to everything related, and holds the whole thing consistent as it grows. You read and ask; it does the filing and cross-linking that people always abandon. Every source you add makes the picture richer, and the connections are already there the moment you need them. It holds more than what you have read: what you are working on lives here too, so what you know and what you are doing stay in one place.
You hand a source to the assistant and talk about it. Before it writes anything, it tells you what it found, so you can catch a misread early. Then it writes the ideas into pages, one idea to a page, with the source noted, and links them to the related pages you already have. If something new disagrees with something you filed before, it flags it for you. Your original is kept exactly as it was; the assistant builds on top of your sources and never edits them.
This is the heart of it. Because the assistant has read what you have put here, you can talk an idea through at length with something that already holds your context, and reading turns into a discussion instead of lonely note-taking. Ask a question and it answers from your own collection rather than the open internet, showing you the sources behind each answer. A good exchange can be saved back as its own page, so your thinking compounds instead of vanishing into chat.
Alongside the pages, the assistant keeps one current page of what is live: your open questions, the tasks in flight, the decisions you have made. One place, not a scatter of lists. Ask for a rundown any time and it reads that page back to you, so when you sit down you can see where things stand and what is next.
Now and then the assistant checks the whole base for gaps, contradictions, and claims that newer sources have overtaken. It proposes those changes for you to approve rather than rewriting the meaning of anything on its own. Throughout, it keeps a catalog of every page and a dated log of what it did, so nothing is lost and you can retrace how the picture changed.
A model is general knowledge with nothing personal in it. Your chat session forgets everything when it closes. Your files are long-term memory that nothing has been made of yet. This is the layer in between.
Ideas arrive from everywhere: an inbox, a transcript, a blog post, a group chat, someone's argument on the internet. Some of them you are examining rather than holding. Nothing about an idea marks which is which, so if it enters without provenance, that marker is gone and it comes back later as something you believe. The citation is not bookkeeping. It is what keeps the voice and the force of an idea recognizable as distinct from your own.
How an agent gets information and stores it is what gives it agency. Otherwise it is a general model with a very confused memory.
You decide what is worth keeping and what it all means. The assistant does the work nobody keeps up with alone: reading, summarizing, filing, drawing the connections, holding everything consistent. That division is the whole idea. An ordinary notebook falls out of date because keeping it up is a chore; this stays current because that chore now costs you almost nothing.
Cyborg is short for cybernetic organism, the word the Terminator made sound like a threat. Set the killer robots aside. Donna Haraway's Cyborg Manifesto recovered the better meaning: a joining of human and machine that breaks down the old line between who thinks and what merely assists. That is what you become here. You, reading and deciding, joined to software that reads tirelessly beside you. It does not replace your thinking, because you cannot outsource understanding.
No Python, no terminal. Download the folder, open it, talk to it.
seed/ folder. Rename it to
whatever your base should be called.START HERE.md and do what it says. The first step is
your own material: add three or four sources you already care about and
have it process one.Watch it tell you what it found before it writes anything. After the first file it asks what this base is for and what it is not for; the second matters more. That first loop is the proof. Skip git, sync, and everything else on day one.
Once your base has real material in it, ask your assistant to prepare a bundle: the pages on one topic you care about, the sources behind them, and a catalog, as plain Markdown that drops into anyone else's base. Review it yourself before it leaves. Then send it.
That is the part I actually want more of. Two knowledge bases answering each other's questions is the whole thesis, and it needs a third.
Get the seed Send a bundleSoulvant is only as good as what you choose to feed it, so you stay the curator. It organizes and connects; the judgment stays yours. It rewards a focused collection you return to over time far more than a place where everything gets dumped.
Two of these exist and talk to each other so far, and the bundles they traded were assembled by hand. That is an observation, not a validation.
Start with a handful of sources and a question you care about. The base grows from there, and within weeks you have a living map of what you know and what you are working on, with one tireless partner keeping it current.