Fine, one model can go down But when every single one is out, nobody is building anything today Go rest
Harmony reportedly had ~4B $ONE issued w/o authorization today To me, an old attack class is becoming relevant again: go after control of the token when attacking the protocol itself is harder A protocol can be technically intact and economically broken at the same time
Opus 5 is getting exhausting I give it one narrow task -> somehow A turns into A+B+C Extra work I never asked for, random context, and sub-agents sent off with huge vague tasks At this point I spend more time managing the model than using it
Seeing more and more safety and DeFi monitoring tools ship. By end of year, DeFi visibility is going to be unrecognizable. Small teams and retail users will be able to see the same amount of data that only the big players could afford before (for a fraction of the cost) BTW big fan of what DefiLlama is building and their vision
Happy to support early-stage DeFi projects with our knowledge, capital and network
Looks like INFINIT is probably hacked 🚨 ~$1.8M lost on BSC (~9M $IN tokens minted and dumped). TX: 0x8d4a72ce9e2f6dea1e9cfc9ff5c19e8d45bf6f4033fcea9739e2409efa3264f2 How it happened: LayerZero cross-chain message forgery. A fake message was sent from Ethereum (EID 30101) to BSC, triggering the INFINIT contract to mint ~9M tokens. All tokens were immediately sold. Root cause: Either the Ethereum-side contract was compromised to send malicious LZ messages, or the BSC contract lacks proper cross-chain sender verification
The @HyperliquidX architecture systematically absorbs adjacent verticals through the zero-cost conversion of its existing active trader base. Sustaining this expansion rate over a six-month horizon mathematically guarantees the protocol will capture the dominant share of prediction markets without incurring external capital acquisition costs.
