Neranu acquires rights to private software, turns it into reproducible agent environments, and licenses ready-made packs for reinforcement learning and evaluation.
A focused collection of executable repair environments covering state recovery, concurrency, parsing, and API compatibility.
Neranu handles source acquisition, rights, task design, verifier hardening, and calibration. Model teams receive a complete pack for training and evaluation.
Pinned containers, reset logic, tools, and agent-ready task interfaces.
Reference solutions, hidden checks, independent review, and baseline results.
A documented rights record and a license scoped to your training or evaluation use.
The founding collection targets reliability failures that public coding benchmarks rarely cover deeply: partial writes, concurrent state, corrupted streams, and compatibility constraints.
Neranu manages sourcing, expert production, validation, and release. Each pack arrives sealed, documented, and ready to run.
Secure rights to private source material with usable technical depth.
Experts turn real failure surfaces into resettable, reward-bearing environments.
Neranu tests determinism, shortcut resistance, provenance, and difficulty.
A frozen pack is licensed for training, evaluation, or managed access.
We are looking for one model team to shape the first collection and its delivery format.