Cloudflare just turned AI agents into SECURITY AUDITORS. And open-sourced the skill. It maps your codebase. Hunts vulnerabilities. Spins up separate agents to VERIFY them. Rejects false findings. Outputs the evidence. Already +3K GitHub stars today.

Cloudflare + Alchemy (@alchemy_run) + Effect (@EffectTS_ ) is such a winning stack with AI agents, it's not even real. Provides pretty much all the primitives you need for building any kind of application and ensures proper guardrails & rules to keep the agent on track. And all the infra lives in code - which is how it should be with AI agents.

@CloudflareDev @Kling_ai @FireworksAI_HQ @vercel @ElevenLabs @stripe @deel 6/ 👨‍⚕️Building the Future of Preventive Health with Jonathan Swerdlin (@swerdlin) of @function Health & @FenwickWest

Cloudflare, on the trillion requests a day it screens: "The question is not whether a determined attacker can get through. They will." Its answer is to raise the price of every attempt until the attack stops paying. Nobody catches every cheat. The trick is making cheating a losing trade. Flop Network puts real money behind that idea. A miner stakes $ FLOP in proportion to the compute it offers, so there is capital behind every session it accepts. Validators check the work certificate and re-run a sample of sessions. A miner caught faking is slashed, up to the whole stake.