Train on real problems
We're an AI Lab building reinforcement learning environments that mirror realistic workflows, using them to train the next generation of agents.
Task Simulation
Realistic environments that go beyond toy problems. Real systems, real complexity, real feedback.
Configurable Chaos
Inject realistic bugs, edge cases, and failure modes. Control difficulty and complexity at every level.
Real Rewards
Dense, meaningful reward signals derived from test results, code quality metrics, and runtime behavior.
RL Integration
First-class support for reinforcement learning pipelines. Gymnasium-compatible API with step, reset, and observe.
Software Tasks Environment
100
Total Tasks
44
Backend Bugs
35
Frontend Bugs
21
Cross-Stack
25
Benchmarks
| ID | Environment | Stack | Difficulty |
|---|---|---|---|
| ENV-001 | be-bug-001 · distance-calc | Py | 0.3 |
| ENV-002 | be-bug-005 · crud-operations | Py | 0.5 |
| ENV-003 | fe-bug-001 · utility-format | JS | 0.2 |
| ENV-004 | fe-bug-012 · stripe-payment | JS | 0.7 |
| ENV-005 | xs-bug-001 · field-mismatch | Py/JS | 0.6 |
| ENV-006 | be-bug-016 · async-patterns | Py | 0.8 |
| ENV-007 | fe-bug-020 · component-logic | JS | 0.4 |
| ENV-008 | be-bug-023 · schema-defs | Py | 0.3 |
| ENV-009 | xs-bug-008 · api-contract | Py/JS | 0.9 |
| ENV-010 | bench-be-003 · service-layer | Py | 0.6 |
| ENV-011 | fe-bug-028 · data-display | JS | 0.4 |
| ENV-012 | be-bug-031 · joins-commits | Py | 0.7 |
| ENV-013 | xs-bug-015 · http-methods | Py/JS | 0.8 |
| ENV-014 | bench-fe-007 · validation | JS | 0.5 |
| ENV-015 | be-bug-042 · route-response | Py | 0.6 |
| ENV-016 | bench-xs-004 · req-compat | Py/JS | 1.0 |
Partners
