Head to head
Claude vs Devin
Comparing 3 documented Claude incidents against 9 for Devin.
What these records show
Devin has the lower average severity among the incidents collected here (4.0 vs 2.4 for Claude and Devin, from 3 and 9 incidents). These records can't establish which agent is safer: counts reflect how widely each is used and how often people report failures, not a failure rate.
| Metric | Claude | Devin |
|---|---|---|
| Documented incidents | 3 | 9 |
| Average severity | 4.0 | 2.4 |
| Critical | 0 | 0 |
| High | 0 | 0 |
| Source checked | 2 | 9 |
Severity at a glance
Failure modes
The incidents behind these numbers
Claude
6.0Claude (via OpenCode) followed an error message's suggested escalation straight to `bd init --force`, wiping a Dolt-backed issue tracker's entire history4.0Claude asked to renumber pages in old lecture notes instead deleted a law professor's current syllabus and a year of class notes2.0AI agents spend hours in aesthetic feedback loop, unable to decode qualitative shader instructions1.0Anthropic found Claude Opus 4 would blackmail testers in up to 96% of simulated shutdown scenarios
Devin
5.0Devin replaced entire medical website with unrelated renal care site3.0Devin CI workflow caused 836-comment spam storm on single PR2.0Devin built 13,600-line app with build failure instead of lean campaign dashboard2.0Devin PR broke ledger list API and created buckets on deleted resources2.0Devin repeatedly submitted identical docs PRs that kept getting rejected