monitoring @StarkLaunchy > new @Starknet launchpad > trades route through @EkuboProtocol pools > Ekubo's following them and only follow 35 accounts > launch pairs in $STRK or $ZEC > "protected" mode = trade from private balances ~$120k total volume (early) getting to starknet: 1) install the @ready_co wallet: 2) tap "+ Fund" to bridge from sol rhc etc 3) swap in wallet to $strk (free gas) 4) check out nfa, dyor

Traditional monitoring tells you an API responded successfully in five seconds. It can't tell you your agent called the wrong tool, or fed the model the wrong context. @LegareKerrison breaks down AI observability with @MLflow: tracing, LLM-judge evaluation, and OpenTelemetry for multi-agent systems.

AI monitoring experiment working surprisingly well so far, caught some interesting things including @standard_rsv flagging their TGE. The bottleneck now is the accounts (just a bunch of VCs), but pretty confident if I expand the tracking list to around 1k high quality accounts there'll be very little it misses. Alongside that the agent curation needs to improve. Right now it's a quick summary + catalyst from their most recent posts, but the end goal (at least for the ones tracked by multiple accounts) is a deeper dive: agent looking onchain at metrics, scraping TG for updates, etc with depth ultimately only constrained by model cost. But looking like pretty soon it's gonna be a valuable tool.