dax rewrote OpenCode Browse in Rust for $60
dax says replacing the parts of Playwright his browser tool needed, in Rust, cost him $60 in tokens. Meanwhile Cursor's security update has Malte Ubl arguing the only real fix is killing long-lived credentials.
Read the whole week in one goFable 5.2 tops the arena and still burns your limit
The big disagreement
Malte Ubl: Cursor's leak proves static credentials have to die
Cursor published an update on its recent security incident with a detailed overview and remediation steps. Malte Ubl's summary is that static credentials leaked through stolen LLM transcripts that contained them, and he says the only real fix is killing long-lived static credentials industry-wide. Cursor's own remediation steps are in its post, not in his.
The only real fix is to kill long-lived static credentials.
Receipts
dax rewrote OpenCode Browse in Rust for $60
dax says he rewrote OpenCode Browse in Rust, including reimplementing the parts of Playwright and Patchright the tool actually used. He puts the total cost at $60. He did not post timings or a diff, so the only number here is the bill.
rewrote OpenCode Browse in rust which included reimplementing the parts of playwrite/patchwrite we needed
How they actually work
Rauch's three steps: feature-gated code, verification gates, agent watches metrics
Guillermo Rauch laid out how he thinks agents will ship all code: the agent writes feature-gated code, verification gates run on push for proof checking and security audits, then the agent monitors metrics. He points at `vercel flags`, `vercel metrics` and `vercel deploy` as the version you can run today.
I haven't touched localhost in month.


