卖什么
We find flaws in RL environments and evals and help you fix them.
商业模式拆解
买家：谁掌握预算？
Teams training models with RL, building environments or running agent evals (labs, eval vendors).
触发点：什么让他们现在购买？
Training amplifies flaws in graders; need to separate real capability gains from exploitation.
工作流：什么输入变成什么输出？
Run an agent swarm (EnvCheck) over RL environment or eval tasks, peer-review findings, report reward hacks and grading bugs to fix; found 110 issues across Terminal-Bench 4.0, DeepSWE 1.1, BFCL v4.
经济账：省了什么，怎么收费？
Tooling and audit service; blog claims audit costs roughly $3.3-$9.4 per task, similar to running the benchmark task.
渠道：怎么接触到客户？
未核实Contact-us led, founder outreach and public research blog.
融资历史与报道
| 融资日期 | 轮次 | 金额 | 投资方 | 来源 |
|---|---|---|---|---|
| — | YC 2026 夏季 | — | Y Combinator | ycombinator.com |
档案汇总自公开报道和公司官网。商业模式部分是 MaxBC 对这些材料的解读，不是已确认的事实；未知之处留空，不作猜测。 更新于 2026-10-12.