ContextQA launches Ship to reproduce bugs before Claude or Codex fixes them
ContextQA's Ship automatically reproduces bugs from Slack, Linear and Jira, documents evidence and finds a likely root cause, then hands the package to Claude or Codex to fix.
When OpenAI built SWE-bench Verified in 2024, human annotators screened 1,699 real GitHub issues. They flagged 38.3% as underspecified. The set that became the standard yardstick for coding agents kept only issues a person could work from. Real bugs skip that filter. They show up as "checkout is broken" in Slack, and somebody has to make the bug happen again before anyone can fix it. Reproduction has always been slow work. As coding agents get faster at writing the fix, it becomes a bigger share of what a bug costs. Google's 2024 DORA report estimated that a 25% rise in AI adoption would bring a 7.2% drop in delivery stability. The 2025 report found AI now lifts throughput and still adds instability. More code ships, and more of it comes back broken. ContextQA launched Ship today to take on that reproduction work. As the company describes it, a bug that lands in Slack, Linear or Jira gets reproduced, documented with evidence and investigated for a likely root cause. From there, an engineer can send the package to Claude or Codex for the fix. In most startups I've seen, nobody owns reproduction. It goes to whoever got pinged first. Hard-to-repeat bugs make that ownership gap expensive. The ones that only show up for one account on one browser can eat days of an engineer's week. They're also the first thing I'd throw at any reproduction tool. Once agents write the fixes, the reproduction becomes the spec they work from.
This post quotes another post: x.com/2548584926/status/2107095904474075536