How do you find a thought from 3 weeks ago?
Most people dig for a bit, fail, and resign themselves to either flying blind or writing better notes next time. With Corpus, you don’t have to do either.
How it works
Just speak. Corpus organizes your brain for you.
You have notes. You have long-running projects.
If there’s time, the what might make it into notes. But the why behind each decision rarely leaves your head. Corpus is where that reasoning goes. Speak to save, messy reasoning and all.
01
Brain dump
Say the thought out loud and move on. Walking out of a call, between meetings, mid-thought. No folders, no tags, nothing to file.
- ◉ 0:19Jul 28“Dana loved the product and then balked at the annual price…”
- ◉ 0:11Aug 4“Marcus wants to keep annual for the ACV. Not convinced that holds above forty…”
- ◉ 0:24Aug 11“Decided. Annual comes out of onboarding. Can’t test two things at once…”
- ◉ 0:13Aug 19“Read the win/loss doc. Two of the three we lost wanted to try before committing…”
67 seconds of talking, across three weeks.
02
Organized automatically
Every person, project and decision gets its own page, and Corpus writes to all of them at once. A single fact lands on every page it is relevant to, instead of staying locked in the one note you happened to say it in.
Your brain map
- Jul 28Dana balked at the price, third this month
- Aug 4Marcus wants it kept for the ACV
- Aug 11Pulled from onboarding, to keep the test clean
- Aug 19Win/loss: 2 of 3 wanted to try first
Open anything on the map and the page is already written. Those four lines came from four separate brain dumps three weeks apart. Each of them is also on Dana’s page, on Marcus’s, and on onboarding’s.
03
Remember any one piece. Corpus connects the dots.
One fragment is enough. Someone pushed back about something, and you cannot recall what. Corpus returns the decision, the reasoning behind it, and what is still unresolved.
All I remember is someone pushing back on something we changed for new signups. What was that, and where did we land?
That was the annual plan coming out of onboarding, on August 11. The reason was measurement rather than price: you couldn’t change onboarding and pricing at the same time and read either cleanly.
The pushback was Marcus Webb, who wanted annual kept for the ACV. What outweighed him: three prospects stalled at the annual price that month, Dana Cho most explicitly, and two of three lost deals in the win/loss doc wanted to try before committing to a year.
Still unresolved: you weren’t convinced annual holds above $40, and you haven’t tested that since.
Answered from 4 memories · Jul 28 – Aug 19
Ownership
A brain you can take with you.
Everything you save exports as plain Markdown — one page per topic, your captures verbatim, the links between them intact. It opens in Obsidian, in Notion, or in a text editor thirty years from now. No proprietary format, no export queue, no plan required.
Where this is going: the same brain, readable by whichever model you prefer to think with. Your memory should not be a feature of somebody else’s assistant.
The part most companies leave out
While Corpus is running, your text is processed on our servers. That means a warrant served on us reaches what we hold, and no amount of marketing language changes it.
What we can promise instead: we do not sell your corpus, we do not train models on it, and we never hold it hostage. Model providers see your text on paid API terms that forbid training on it — not the consumer terms that permit it.
Pricing
Free while Corpus is in beta.
Later, asking your brain questions becomes a subscription. That is the part that costs real money to run, and charging for it is what keeps the rest of this honest — there is no version of this product where the business depends on doing something else with your corpus.
Reading and exporting everything you have saved stays free, permanently. What you wrote is yours whether or not you are paying us.
Questions
The ones worth asking first.
- Is this for me?
One question sorts it: if you took two weeks off, would you come back to a new set of problems, or the same ones where you left them?
Corpus is built for the second answer. Long projects, where today’s decision depends on context you gathered weeks ago and the reasoning is still only in your head. If your work resets daily, a good assistant is enough and this is more machinery than you need.
- Why not just use ChatGPT or Claude’s memory?
Because of what they can see. Their memory is fed by one channel: the things you typed into that chat. And you only ever type what you needed to say to get an answer, so everything else you know never reaches it.
What they keep is also a profile rather than a record — a synthesized description of you, updated in the background, not the specific thing you said on a Tuesday in July and the reason you said it. Corpus keeps your words.
- Why not Notion or Obsidian?
They are excellent, and they make a trade Corpus refuses: better retrieval bought by taxing capture. Backlinks, tags and structure work well, and every one of them is your job at the moment you write.
That tax never falls, so the discipline lapses, and the retrieval you were paying for degrades while the cost stays. Corpus inverts it. You file nothing, and the structure is built for you.
- Why not put this in our team wiki?
Some of it should go there. But a shared surface only ever holds what you are willing to say in public, and it has to look finished before you will post it.
Your honest read on how a negotiation is going, or why you think a decision is wrong because of what happened at your last company, has no home in Confluence. That is exactly the context this is for, which is why one Corpus belongs to one person.
- Why not the notebook I already keep?
The notebook is the real competition, and it wins on capture: nothing is faster or more trusted than a page and a pen.
It just never tells you anything back. Six weeks later you know you wrote it down and you cannot find it. Corpus is trying to match the notebook on the way in, by letting you talk, and beat it completely on the way out.