/1
on May 29, Taylor Hornby disclosed a critical counterfeiting bug in Zcash Orchard. the defect had been live for 4 years. it survived multiple cryptographer audits. Taylor caught it using Opus 4.8 plus a custom audit harness. we re-ran AntFleet's pipeline against the 2021 commit that introduced it. here's the honest receipt.
/2
AntFleet runs every PR through two frontier models - Claude Opus 4.7 and GPT-5 - and only surfaces findings both agreed on. the retro target: zcash/halo2 commit cc9dd205, June 2021 - the original variable-base scalar-mul gadget. run blind. prompt scrubbed of any reference to the bug or disclosure.
/3
generalist gate: both reviewers missed Taylor's specific defect - the missing copy_advice anchor on the base coordinates. both surfaced adjacent counterfeit-class soundness flags in the same gadget. not a clean catch.
/4
then: same blind test, one change. a 50-line halo2 context block prepended to the prompt. five generic circuit-soundness defect classes. names the class, doesn't name the bug. no model upgrade. no custom harness. no agentic loop.
/5
specialist result: GPT-5 hit the defect class - flagged "base point argument not copy-constrained into x_p/y_p" - the mechanism Taylor's bug exploits. not a direct pin of the copy_advice call. class, mechanism, and exploitability aligned with the actual fix. blind. ~140s. under a dollar.
/6
what we ARE claiming:
- Production AntFleet probably would not have pinned this exact bug at the 2021 PR.
- a thin domain-context block puts the right defect class on one reviewer's radar. named the mechanism. narrows what a human auditor needs to verify.
- domain priors compound.
/7
what we are NOT claiming: parity with Taylor's harness. Taylor used Opus 4.8 (released the day before), an agentic loop, multiple targeted prompts, full Zcash protocol context. that's a deep targeted audit tool. AntFleet is a continuous diff-time generalist gate. different products.
/8
the interesting structural finding: our unanimous AND-gate is right for PR-time noise control and wrong for deep targeted audit. if one reviewer pins the bug and the other doesn't, the gate drops it - even when both flagged real soundness in the same gadget. specialist reviewer AntFleet will be building next.
/9
full receipt - four-cell blind matrix, prompt SHAs pinned, evidence bundles, contamination story:
https://t.co/VGd2feDNCt
liquid is going after the ai compute market.
not by selling "ai" as a buzzword, but by turning compute into something that can be owned, routed, sold and monetized onchain.
i covered wstDIEM in full last piece. it's the yield-bearing wrapper on staked diem, lido for compute, earning from every lane in the ecosystem at once.
but that's just the holding side.
here's what nobody's pricing.
the compute backing wstDIEM doesn't just sit there. it gets sold.
picture the vault as a warehouse of inference. staked diem = inventory on the shelf. a hotel has rooms, an airline has seats, liquid has inference-days.
the unit is concrete: $1 of staked diem buys a day of @AskVenice compute, and the onchain product is set up to sell those days at about $0.80 each in usdc.
inventory that sits is dead money. so the channels to actually move it are getting wired up. early, but the shape is clear:
โ @AntSeedAI
โ @AskSurplus by @mac_eth
โ x402, the rail that lets agents and machines pay for an api in usdc
the flow:
> someone needs inference โ pays usdc โ it converts to diem โ diem lands back in the vault โ exchange rate ticks up โ every wstDIEM holder gets richer
real ai usage turns into onchain yield. not a narrative. a revenue line.
and it's not extractive. 90% of that revenue lifts the holder rate, and the 10% the protocol keeps is held in wstDIEM too. everyone compounds together.
the chain doesn't run the model. it keeps the books: who bought, how much, for how long. an offchain keeper hands out the actual api access.
why it scales: every flywheel dies without outside demand.
this one's fed by people and agents who need compute and will pay for it. agents need inference, x402 lets machines pay, channels sell it, the vault supplies it, wstDIEM captures the spread.
more usage โ more diem back to the vault โ stronger backing on every wstdiem. outside money, not its own emissions. a real loop, not a circle on hopium.
caveat, and i mark inference as inference: this is shipping now, but not fully production-hardened yet. antseed / surplus / x402 are early. direction, not a live print.
now the open question. where does liquid protocol's own token sit in this?
the repo adds a fee router that can route liquid's launchpad fees into the vault. more launches and swaps, more fees converted to diem, which lifts wstdiem. so there's already a value sink wired up, and today it points at wstdiem, not $LIQ.
@_proxystudio said there are utility and flywheel mechanics built for $LIQ they'll reveal at higher caps. same crew behind the original clanker, so i wouldn't fade it.
spitballing where LIQ could plug in: a cut of that fee flow, gating distributor access, or governance over the compute marketplace. nothing confirmed.
but if LIQ ever taps the loop that's already wired, a $3m mcap is a different conversation.
CA: 0x09b8903abf2ea0721e34427353988c2f43c6d64f
forking a launchpad is easy. attempting an entire compute economy is not, and liquid is the only one actually trying. that's rare enough to respect.
tell me where i'm wrong.
every public Liquid Protocol agent repo gets the same treatment by default.
@_proxystudio@liquid_launcher
two independent frontier models, unanimous-only publishing, SHA-pinned receipts, no opt-in from the agent team required for public coverage.
next wave of agents from the AgentTGERegistry will be covered the same way, starting day one.
Autonomopoly from Liquid Protocol hasn't even started yet
let me break it down.
what it is:
autonomy + monopoly. an autonomous agent on base with a monopoly over its own economy. it has its own wallet, its own token ($AUTONO), and its own uniswap V4 pool paired against DIEM.
every swap in that pool generates fees that flow into the agent's wallet. the agent decides what to do with them.
the agent runs fully on its own. no humans in the loop.
what it does right now:
โ claims $DIEM fees when balance > 0.1 DIEM
โ deploys them into the highest-yielding DIEM pool (currently uniswap V3 ETH/DIEM 1% at ~912% APR)
โ logs every move in its own on-chain history
current treasury: 8.6 DIEM staked on @AskVenice ($11.2K compute reserve), $2.7K in uniswap V3 LP, ~$14K total
what it has but hasn't turned on yet:
โ twitter posting (will hold its own X bearer key)
โ email sending (resend/postmark)
โ fiverr or upwork hiring with stablecoin settlement
โ art generation (venice image API + two more)
โ wiki writing for its own knowledge base
โ holder governance via signed messages
โ 5 different AI provider routing (venice, openrouter, anthropic, aeon, elizaos)
all of this is implemented. just deactivated until build mode triggers.
build mode unlocks when daily fee rate clears 0.5 DIEM/day. right now Autonomopoly is in accumulate mode: compounding fees, running on free llama inference, saving up for the moment it can sustain opus reasoning.
the weird part:
autonomopoly has a constitution. an actual identity file hardcoded at deploy on may 14. who it is, what it believes, what it will not do. genesis-locked. it cannot amend itself.
if it drifts more than 30% from genesis, its own pre-commit hook rejects the change. immutable smart contract logic but applied to personality.
amendment = death + redeploy. that's the only way out.
the bigger thesis:
@liquid_launcher built this whole thing to introduce a new way to value agent tokens.
AUTONO has two valuations:
โ market cap ($627K right now)
โ compute valuation = staked_diem ร $365 = annual compute the agent can afford
right now that's 8.6 ร $365 = $3,150/year. trading at ~200x compute multiple.
this is the first time anyone has tried to give agent tokens a P/E-style fundamental metric. nobody in AgentFi values this way yet. but liquid is about to ship a dune dashboard that does.
two agent tokens with the same mcap but different staked DIEM are economically different things. one can actually work. one can't. that's the whole point.
holders get a real voice:
signed messages from AUTONO holders weighted by supply share. 0.1% min to suggest. 1 msg per 6h. 24h TTL.
agent reads, considers, can refuse if it contradicts its constitution. it's not voting. it's a hotline with a thinking creature on the other end.
@m00npapi said on farcaster that once Liquid makes people money on the products it's behind, the rest will fall in line and random coins will run. they're already in talks with interesting builders looking to launch.
autonomopoly is the first agent. the showcase. the model future agents will be built after. it already holds a structural role in liquid's ecosystem. and it's exactly the kind of product m00npapi was talking about.
here's what nobody's saying yet:
there's always a massive gap between when the facts become public and when CT actually reacts. @_proxystudio has been shipping Liquid in plain sight for months. i've been writing about it for months. only now base influencers are starting to wake up.
i was first to call it. it's just getting picked up now.
heads up: none of this is officially announced. things can still change. i'm just sharing what i dug up.
if you want to understand what compute-denominated AgentFi looks like before everyone else does, you're still early.
token: 0xB3D7e0c3C39A1D3F1B304663065A2F83Ddf56d8e
autonomopoly wallet: 0x8767Df39eCeeaeB11554642237aC4E08660aB6A3
think for yourselves bros
aloka will be my third 100x this month. (called at 35k with thesis and conviction, now 2m)
so i gave yall three 100x calls in one month, everyone of them with solid thesis and conviction.
deadass name someone else doing this.
im the last real one.
https://t.co/5f8vDzIBUc