The Clasp: All Your Best Selves at Once

For about eight months I’ve been running the same loop: record a thought into a voice memo, hand it to an AI, turn it into a conversation, index that through a wiki, compile it into an ebook and a framework, and reuse the whole lot in the next conversation. For most of that time, if you’d asked me what I was building, I would have pointed at the ebooks. That was the mistake. The ebooks were never the asset.

The asset is the thought. It’s my thoughts, not my ebooks. And that correction — which sounds like a small change of emphasis — changes what kind of thing you’re accumulating, what it costs to add to it, and why the returns feel like they’re accelerating rather than merely adding up.

This is an argument about the unit. Not how fast the loop spins, not how you learn from it — those are their own stories. Just this: what, precisely, is the durable asset, what does it now cost to secure one, and what happens when you have a great many of them present at once.

The correction: the asset is the thought

Not the ebooks. Not the frameworks.

A thought, unaided, is the most perishable asset a person owns — half-life measured in hours. The traditional fix is to write it down. But writing it down is lossier than we admit. Go back to something you wrote three years ago and you’re reading someone else’s text. You can cite it. You can’t re-enter it. The reasoning that produced it is gone, and if you want to extend the idea you have to re-derive your way back in.

Mathematicians know this feeling exactly: you understood the proof once, you lost it, and now you must climb the hill again.¹ The written artefact preserved the conclusion and threw away the ability to think from inside it.

So when I say the asset is the thought, I’m claiming something stronger than “I kept my notes.” I’m claiming the thought itself comes back at working fidelity — re-enterable, extensible, still warm — not a fossil of it. The ebook was never the asset. It was the compiled binary. The thought is the source, and I kept the source.

The ebook was never the asset; it was the compiled binary. The thought is the source, and you kept the source.

The clasp: securing collapsed into speaking

A named economic event

Once I clasp the idea, I’ve got it to reference forever. Clasp is the right verb, and it names an economic change most people haven’t noticed yet.

There used to be a gap between grasping an idea and securing it. To keep a thought you had to stop, write, formalise — expensive enough that most graspable ideas were simply dropped. A thought you didn’t develop immediately was a dead loss. That gap is what rationed thinking.

The pipeline collapsed that gap. It folded securing into speaking. The marginal cost of permanent, retrievable, provenance-carrying tenure for a thought is now the cost of saying it aloud into a voice memo. Everything downstream — the formalising, the edges, the indexing — happens on its own.

Here is the part where I’d say I knew it before I knew it: the architecture already agreed. The wiki’s unit was never the document — it’s the claim with edges, which is exactly turn-sized. A claim with edges is a thought, formalised: one assertion plus how it fits. That’s precisely why the system can hand back thoughts instead of artefacts. Document-grain systems — RAG, SharePoint, every knowledge base ever built — can only return the containers thoughts were buried in, and leave the exhumation to you. Make the thought the addressable unit, and the thought is what’s addressable.

What you can now afford

Fragments, forks, parked half-ideas

Change the cost of securing a thought and you change thinking behaviour, not just storage. When capturing a fragment is free and losing it is impossible, you can afford to think differently:

You can afford fragmentary thoughts — the fumbling half-thought that isn’t worth writing an essay about, but is worth keeping. You can fork freely, chasing a tangent without committing to it. You can park half-ideas without grief, because nothing parked is lost — and parked things come back unsolicited, at the right moment, wearing their provenance.

My own corpus already named the payoff: amortised thinking — comprehension paid once at ingest, interest collected on every future conversation. The refinement is what the interest is paid in. Not documents. Thoughts — yours, at their best, forever callable.

The room where thoughts meet

A capability that didn’t exist

Now the part that actually matters, and the part I kept fumbling toward while the machine got described back to me. Nobody in history has been able to do this. Not “no one had a wiki” — no thinking person, ever, has had their own accumulated thoughts co-present at the moment of a new problem.

Unaided thought is serial and lossy. Working memory holds only about four items at a time,² and only the recently-rehearsed ones show up uninvited. Your best March idea and your best May idea only ever meet if both happen to surface in the same moment — and across a lifetime of thinking, almost none of your ideas ever meet each other.

That’s the silent tragedy of ordinary cognition: a mind full of insights that never got introduced.

So when I bring back seventeen thoughts I’ve had over several months to bear on the same problem, at the same point in time, that isn’t better recall. It’s a capability that didn’t exist. The word for it isn’t “remembering.” And the reason the word matters is arithmetic.

A hundred and thirty-six collisions

Combinatorial, not additive

Seventeen thoughts on one problem is not seventeen units of value. It’s a hundred and thirty-six possible collisions between them, available right then — because seventeen things pair up in 17 × 16 ÷ 2 = 136 ways. The value was never in the thoughts stored. It’s in the meetings between them.

That last number is the whole point. The eighteenth thought you clasp isn’t worth one more unit. It’s worth seventeen new collisions with everything already in the room. Each thought you add doesn’t raise the total so much as raise the rate. That is why the returns feel like they’re accelerating eight months in — they are. Additive accumulation grows in a straight line; the meetings grow as a square.

This is why “compounding” is exactly the right word and “remembering” isn’t. A memory aid gives you back the average of what you once thought. This gives you the meetings between your maxima. You built the room where they all finally meet.

The return path is half the mechanism

Capture is only the first stroke

None of this works if the thoughts just pile up somewhere safe. Capture without timely, working-fidelity return is just an archive — and archives are where thoughts go to be technically-not-lost while remaining practically unreachable.

The return has to land inside the life of the problem that summoned it, at a fidelity you can actually re-enter, pointed at the thing in front of you. When it does, an old half-thought arrives already fleshed out — more than it was when you first fumbled it, because it’s been elaborated in the meantime — and it arrives because the current problem reached for it, not because you went digging. That’s the difference between a warehouse and a colleague.

All your best selves at once

The possessive is the point

I’ve said “my” enough times now that it deserves to be honoured directly. Nothing in the chain alienates the thought. It’s my thought at the encounter, my thought through the elaboration, my thought I learn on the read-back, my thoughts converging on the problem. It never becomes “the wiki’s content” or “what the AI knows.”

Which means the thing that’s been compounding for eight months isn’t a knowledge base I happen to own. It’s me. A person is normally only ever their current self — today’s draw, plus a dim memory of who they were in March. I’ve arranged to be all of my prior best selves at once, on demand, pointed at whatever’s in front of me.

Seventeen thoughts brought to bear on one problem is seventeen versions of you, each at the moment they were sharpest, sitting down at the same table.

That’s the finer point, and it doesn’t need the mechanics. You can build the machine — the voice memos, the conversations, the wiki, the edges, the compiled binaries. But the machine is only ever the how. The what is quieter, and it’s the only part that walks out of the room with you.

The machine is how. You, accumulated — that’s what.