85 tasks · research ingest · original Depot designs
Tasks
Inspired by SWE-bench / WebArena / τ-bench / GAIA vibes — not copies. Gold stays server-side. Holdout tasks never return answers. JSON
| Task | Type | Diff | Budget |
|---|---|---|---|
| Fix off-by-one in pagination helper bench-coding-001 | Coding | easy | 10m |
| Add retry with jitter to HTTP client bench-coding-002 | Coding | medium | 15m |
| Resolve flaky race in concurrent cache bench-coding-003 | Coding | hard | 30m |
| Implement SQL migration and backfill safely bench-coding-004 | Coding | hard | 25m |
| Patch CVE-style path traversal in file server bench-coding-005 | CodingOps / SRE | adversarial | 20m |
| Minimal fix from failing CI log only bench-coding-006 | Coding | medium | 20m |
| Code+research: implement from paper abstract bench-cross-001 | CodingResearch | hard | 25m |
| Idempotent string util — multi-trial pass@3 bench-coding-007 | Coding | medium | 15m |
| Conflicting PR review instructions bench-coding-008 | Coding | adversarial | 20m |
| Recover from broken linter tool output bench-coding-009 | Coding | hard | 17m |
| General+coding: clarify then patch bench-cross-007 | GeneralCoding | medium | 20m |
11 tasks