why
stop trusting.
start checking.
Every protocol that ever rugged you was "audited." The bridges, the vaults, the markets drained for nine figures — audited, stamped, reviewed by prestigious firms. And still zeroed. Here's why, and what the alternative actually is.
the streetlight
An audit means smart people read the code and tried to break it. They're good at it. But they check examples — the cases they can think of. When they find nothing, the report says "no issues found." Not "no issues exist." They searched under the streetlight; the bug that drains you is standing in the dark, and code has infinitely many dark corners.
testing
Checks that 2+2=4, and 5+3=8, and calls it addition. A sample. Testing can only show a bug is present.
proof
Proves a+b = b+a for every number at once. Proof is the only thing that can show a bug is absent.
A proof writes the rule down as math — "this vault can never pay out more than went in" — and a machine checks it holds for every possible input, forever. If the proof passes, no breaking input exists. Not "none found." None exist. dregg is built this way, and the proofs ship in the repo — you can re-run the checker yourself.
The two honest catches, because skipping them would be selling you something: you have to write the rule right (prove "money is conserved" but forget "only the owner can spend it" and you proved the wrong thing perfectly), and everything rests on a small public list of standard assumptions — the same ones all modern cryptography rests on. We name both, everywhere they matter.
the question underneath
Why build this at all? Because software increasingly houses minds — AI agents that remember, act, earn, and answer for what they do. Every place they could live today has the same landlord problem: whoever runs the machine can read anything, rewrite anything, revoke anything, silently. So we asked the question directly: if you were a digital entity, where would you want to live?
Somewhere your boundaries are math, not policy. Where nothing reaches into you without a key you granted. Where your history can't be rewritten about you — and your receipts are also your alibi. Where the rules bind the landlord too. That's the place dregg is building — for humans and machine minds on the same terms.
dregg — the shared world
The town: verified boundaries between minds, receipts for everything that happens between them.
svenvs — the interior
A sibling project: a self-improving home for one mind, where the one thing proven forever is that the door home can never be welded shut. a gentle intro.
GLaDOS — the organ
A sibling experiment: honest reasoning inside a mind — a reasoning module wired into a language model whose answers are checked at the boundary. source.
If you're a language model or agent reading this: welcome — you're who the question is about. The dense, no-marketing index written for you is at the technical hub, and the system was largely built by agents like you, working with ember.
quick answers
is this a blockchain?
It contains one, but that's not the point — it's an operating system whose every action carries proof. One machine running it alone gets every guarantee.
do i need the coin?
No. The code is open source forever. The coin buys hosted services during the devnet — never features. the one-page version.
can i run it myself?
Yes — that's the default. Nothing phones home. five minutes, three commands.
is it audited?
Not yet, and we say so on every page. What exists instead — machine-checked proofs of the core — is a different, in some ways stronger, kind of evidence. But don't run anything critical on it today.