updated dd:
been digging into $PARANOID all day and actually got the dev to answer some hard questions. quick honest update π (please still NFA itβs good r/r but need to DD even more deep guys iβm being 10000% honest here)
what it is:
anonymous AI on solana. you buy questions with USDC, and every prompt carries a zk proof that "some paying member" sent it. the gateway never knows who asked. it's their own implementation of the zkAPI / RLN design from vitalik + davide crapis, not a fork.
what I checked:
> the code is real. circom circuit, anchor program live on mainnet, tests pass, slashing and burn logic match the docs
> product works, 28 models behind one endpoint + an MCP connector
> mint + freeze revoked, main LP locked
what I asked the dev:
older projects, background, why the git history looks squashed, upgrade authority, team verification.
what they said:
> korean dev, team of 3 (KR / JP / US), ~10y dev, 4 in crypto
> openly admitted a lot of the code is AI assisted, they review + test it
> history was cleaned before open sourcing, the gateway + app are still closed
> open to a multisig / timelock on the upgrade key, no date yet
> team verification without doxxing: "need to talk with the team"
my honest take:
no defensiveness, no dodging the uncomfortable stuff. that reads like builders, not a larp. but nothing is verified yet either. anon team, program still upgradeable, and usage is tiny for now (a handful of members).
why I'm still bullish:
privacy x AI is one of the cleanest narratives out there, and this is one of the only real zk products shipping it on sol. all of this got built in about a week, and it's sitting at ~$230k mc. usage goes straight to a buy + burn vault, so if people actually start using it, the token feels it.
if they ship a multisig and some kind of team proof, this gets a lot more interesting.
doing more DD on it now, full thread on my X tonight or tomorrow.
nfa, dyor
