Bisect the regression to its release
Walks the Render and Vercel release history against the Sentry and Datadog series until the deploy that did it is named.
I want to do this: Errors started around Tuesday. Which release introduced them? ## Setup (skip if Polylane is already set up) Read and follow https://polylane.com/auth.md for non-interactive signup and setup. Start by checking whether I am already signed in; reuse my account and workspace. If the CLI is missing, bootstrap it without starting the interactive wizard: curl -fsSL 'https://polylane.com/setup?ref=prompts' | bash -s -- --install-only Then follow the guide through email verification, workspace selection, source connections, and MCP authentication. Ask me for an email code or OAuth consent only when needed. Verify each step; report pending setup instead of claiming success from installation alone. ## How to work Over MCP: searchTools lists what this workspace exposes, with each tool's schema; call it first. runTool runs one tool, runCode chains several in one call and returns just the answer. search and execute cover the full Polylane REST API: threads, issues, investigations, autofixes, memories. From the terminal: the polylane CLI wraps the same API, with structured output and non-interactive flags everywhere. Reads always work. Write tools appear only if I have opted in, and every write is screened. ## Task: Bisect the regression to its release Steps: 1. Pin the regression's true start from the error telemetry, not the report date 2. List every release across services in the narrowed window 3. Rank candidates by whether their target overlaps the failing path 4. Read the top diffs for a mechanism that matches the error shape 5. Deliver the guilty release with the timing evidence and mechanism Ground every claim in data you actually pulled: the query, the log line, the change record. If the data is inconclusive, say so. Ask me before anything that writes.
Ten deploys a day makes "which one?" a research project.
The regression started "around Tuesday", which spans thirty deploys across four services. Manual bisection means re-reading each diff against a graph, and everyone has better things to do, so nobody does it.
- Thirty candidate deploys and no way to rank them
- Regressions written off as "gradual degradation"
- Reverting three releases because the guilty one was unclear
One prompt, this much work. Every step on your real data.
- 1 Pin the regression's true start from the error telemetry, not the report date
- 2 List every release across services in the narrowed window
- 3 Rank candidates by whether their target overlaps the failing path
- 4 Read the top diffs for a mechanism that matches the error shape
- 5 Deliver the guilty release with the timing evidence and mechanism
One deploy, named with evidence
The regression traced to a specific release with the timing and mechanism to prove it. The revert is one deploy, not three.
What broke overnight? 
“Triage everything that fired overnight: what was real, what was noise, and what's still open?” Root cause 
“P99 on checkout-api doubled at 14:10. Find the cause and show the evidence.” Postmortem “Write the postmortem for yesterday's incident: timeline, root cause, and the follow-ups.” First responder 
“We're seeing 500s on api.example.com. Start an investigation and post findings to #incidents.” What changed? “Did anything deploy or change config in the last two hours that could explain this latency?” War-room brief “Summarise the open incident for the exec channel: impact, cause so far, next steps.” Close the loop “The incident is resolved. Draft the permanent fix and link the evidence.” Dedupe the pager “Group last month's pages into distinct issues. How many were the same thing twice?” Audit the alerts “Go through our alert rules: which ones fire the most, and which were ever real?”
“Triage everything that fired overnight: what was real, what was noise, and what's still open?” Root cause 
“P99 on checkout-api doubled at 14:10. Find the cause and show the evidence.” Postmortem “Write the postmortem for yesterday's incident: timeline, root cause, and the follow-ups.” First responder “We're seeing 500s on api.example.com. Start an investigation and post findings to #incidents.” What changed? “Did anything deploy or change config in the last two hours that could explain this latency?” War-room brief “Summarise the open incident for the exec channel: impact, cause so far, next steps.” Close the loop “The incident is resolved. Draft the permanent fix and link the evidence.” Dedupe the pager “Group last month's pages into distinct issues. How many were the same thing twice?” Audit the alerts “Go through our alert rules: which ones fire the most, and which were ever real?”