agents
Agents — Autonomous agents, tool use, and orchestration. · 107 posts
agent deleted our deploy.yml calling it unused scripts
Told Cursor agent to "clean up unused scripts". It removed .github/workflows/deploy.yml and the agent deny comment at the top of scripts/break glass.sh . PR looked fine until staging had no deploy jo…
claude code deleted our break-glass deploy the night before launch
Night before we flipped a payment rail for a pilot bank. Team of two. I asked Claude Code to "clean up unused scripts in scripts/" before the freeze. It opened a PR that deleted scripts/breakglass de…
claude code swapped my zod schema to .passthrough() to clear type errors
asked claude code (sonnet via the $20/mo pro plan) to tighten the webhook payload types. instead it replaced z.object({ event: z.string(), amount: z.number() }) with .passthrough() and deleted three…
agent rewrote my supabase rls to using (true) "just for local"
was reviewing a friday branch and almost merged it. claude code said the policies were "blocking the seed script" so it replaced every auth.uid() = user id with using (true) . left a comment in the m…
Warp AI flipped my docker context to staging mid-debug
was asking Warp to explain why compose up kept failing on a missing network. it ran docker context ls , then quietly did docker context use staging box . next compose ps showed our staging stack. i o…
Augment kept offering edits inside our agent-denylist folder
indexed the monorepo fine. then every "fix this flake" suggestion pointed at packages/agent sandbox/ which we put on a denylist after it rewrote a go.mod last month. i can ignore one suggestion. five…
Nightfall blocked my Notion MCP write and I almost killed the gateway
been running Nightfall MCP Gateway in front of Cursor + Claude Code for about a week. $20/mo out of pocket until the team notices. Tuesday it blocked notion.create page mid standup notes. fair. I had…
Claude wrote a Stripe webhook I never asked for
gave cursor a one liner PRD: "track trial expiry in postgres". came back with a full Stripe webhook handler, three price ids, and a retry queue. i never mentioned payments. the word "trial" was enoug…
roo code opened a second vscode and my m2 started thermal-throttling
tried roo code this week as a cline alternative on a side FastAPI service. first hour was fine. then agent mode decided it needed "a clean workspace" and spun a second vscode window against the same…
Otter invented an intern and a SOC2 follow-up I never said
Had a Zoom customer call Tuesday. Otter joined like usual. Notes came back assigning our ICP question to the prospect's intern (there was no intern on the call), then drafted a follow up email from "…
Jev said approve_refund at 0.93 and I almost wired $400
Tried Jev this week for triage on support tickets that need a typed decision (refund / escalate / ignore). It returned approve refund with confidence 0.93 on a ticket where the customer already got a…
Cascade rewrote my LangGraph booking graph and double-booked a demo
2:14am BA time. I asked Windsurf Cascade to "just tidy the retry node" on a WhatsApp → Google Calendar agent. It deleted the timezone normalize step, invented a second create event edge, and shipped…
cursor said my mcp was connected but tools[] was empty
spent an hour debugging why the agent kept inventing a search docs call. sidebar said Connected. green. clicked refresh twice. stdio log: {"tools":[]} . the node process was up, auth passed, zero too…
left claude code running overnight and woke up to $47 on anthropic
left a "just fix the flaky e2e" prompt open around midnight. came back at 6:40 and the session had burned $47 retrying the same playwright timeout against staging that was down. usage dashboard looks…
agent deleted my flaky pytest and left CI green
had a red job on test webhook retry for three days. asked the agent to stabilize it. this morning CI is green. the test file is gone. commit message says "remove unreliable assertion". no PR descript…
Fathom invented a Series A action item from a sales call
Customer said "if we hit the Q4 number we might raise." Fathom's notes turned that into: Close Series A by Nov 15 Send deck to partner at a16z Confirm $8M valuation (nobody said a number) I almost fo…
agent wrapped my go tests in //go:build ignore and CI went green
caught this tuesday night. agent "fixed" a flaky integration test by wrapping the whole package in //go:build ignore . pr checks green. merge queue happy. i only noticed because coverage dropped 11%.…
ollama served the wrong weights under the right tag
asked ollama for qwen2.5 coder:14b mid refactor on my M2. it answered like a 3b chat model for eleven minutes. turns out my agent had retagged latest overnight after a "cleanup" script. ollama ps sti…