Reading the existing arc perp DEX's OrderTypes.sol while writing mine.
Same Order struct. Same EIP-712 typehash structure.
Different domain string on purpose: "ArcPerpRef" / "v1" instead of theirs.
If two contracts share an EIP-712 domain, signed orders become replayable between them. Easy to miss when you're forking someone's pattern. Mine's intentionally distinct so the matcher only accepts what's actually meant for it.
Building a permissionless perp DEX for Arc.
there's now a formal proof that your agent's vector memory forgets what you stored and fabricates things you never did.
scaling it up makes both worse, not better. "the price of meaning" (arxiv 2603.27116) proves it for any memory that retrieves by similarity in an embedding space.
the same geometry that lets embeddings generalize creates competitor mass in every neighborhood.
add data and the crowding grows: retention decays toward zero, and false recall can't be tuned out without throwing away true hits.
not a bug in your pipeline. the shape of the math.
i learned this the expensive way. 500 stored facts, two weeks into a build, a user asks what i know about their job.
retrieval hands back four fragments from different weeks:
"i love my job," "thinking of quitting," "my manager is supportive," "my manager micromanages."
the agent invents a clean synthesis of all four. the user had switched jobs in between. embeddings measure similarity, not truth.
the topology angle says stop storing meaning as geometry, store it as structure.
navigate an edge in an AST or a graph instead of searching a neighborhood.
no crowding, no decay, no false recall.
FORGE backs it: plain AST checks catch structural hallucinations at 100% precision (arxiv 2601.19106).
here's what the structural pitch skips. the same proof shows pure structure escapes the geometry only by surrendering the connections embeddings find.
you trade fabrication for blindness.
so i stopped picking a side. what actually ships across thousands of sessions:
β extract facts, not transcripts
β resolve conflicts on write: archive the old job, mark the new one active
β hybrid retrieval: vectors for discovery, graph for precision
β decay plus nightly consolidation, so memory keeps what matters and lets the rest go
memory is infrastructure, not a feature. that reframe is the whole game.
and it's being measured now.
WorldMemArena (may 28, arxiv 2605.29341) scores these paradigms head to head, embedding memory against retrieval-augmented against terminal-agent harnesses, across multimodal action-world tasks. the question moved from "does it remember" to "what kind of memory survives scale."
full architecture, with the code, here:
I've the same laptop and am frustrated
It's still under warranty need to get body repaired as it cracked because of how messed up the charging port was
it used to get so hot that it burned my hand many times
And now the body is cracked and bent by how hot it got
And the support is so slow to do anything, I can't give them my PC for a whole week but that's the earliest they can give it to me by π₯²
Tangent DEX update
We renamed the repo to it's proper project name:
https://t.co/1nLxlpN3cS
The old link should still redirect to this, so nothing breaks for anyone
Tangent has moved past the initial Arc primitives into a fuller perp DEX reference with:
- Permissionless accounts
- USDC collateral
- Signed EIP-712 orders
- Deterministic matching
- Settlement
- Margin checks
- Withdrawl health checks
- and liquidation paths.
Still early, but the goal is clear: a forkable Arc-native perp DEX reference that builders can study and extend without permissioned matching or custodied account setup.
Possibly available as an API if required
DMs are open for any discussions