gpt-6POST
Astra distrusts subagents and re-verifies their results, one user says
A user reports that GPT-6 Astra re-checks every subagent result even when the conclusion is sound and sources are reliable, wasting tokens.
KC
@ScarletKcArchived 2026.09.14Originally posted 2026.09.14
KC @ScarletKcOriginally posted 2026.09.14
我真的服了，GPT-6 Astra 完全不信任 subagent，返回什么结果它都要自己重新验证一遍 明明对方已经给了合理的结论，可靠的来源也附上了，它还是得亲自再查一次。那你把任务分出去干嘛，最后还得自己重做，真的浪费 token 用下来感觉它太严格了，限制和禁令感特别重，连自己叫来的 subagent 都信不过
x.com/scarletkc/status/2099270103191974280Archived 2026.09.14 15:56