ChatGPT Pro is getting insane. 10 subagents in a normal chat, sessions lasting up to 5 hours, and a massive productivity boost.
retained text
I've been testing the latest ChatGPT Pro updates pretty heavily, and I'm genuinely impressed with how much the workflow has improved. I'm now seeing **up to 10 subagents being used within a regular ChatGPT conversation**, with some of my sessions running for up to 5 hours while working through complex engineering tasks. For context, I've already gone through **over 1.5 trillion tokens in Codex**, mainly working on blue-team security, vulnerability detection, code auditing, security testing, bug hunting, fixing issues, and regression testing. A lot of my work involves large codebases where a single task can require hundreds or thousands of checks. The interesting part is seeing multiple subagents work on different areas, investigate problems, implement fixes, run tests, review each other's work, and continue iterating. I've had sessions where the system was simultaneously working on security engines, backend logic, API integrations, test infrastructure, reports, and independent verification. It's not perfect, of course. Sometimes fixes introduce new regressions, tests fail, or the agents need another iteration. But that's actually what makes these longer sessions interesting. The model can keep working through those failures instead of simply stopping after the first attempt. The ability to delegate parts of a complex task and maintain progress over such long sessions feels like a significant upgrade. I still verify the results and wouldn't treat a successful test run as proof that everything is secure. But from my experience, this update has the potential to speed up serious development and security work by a huge amount. I'm curious whether anyone else using Pro has noticed the same thing. **Have you tried the new subagent workflows in regular ChatGPT chats? What's the longest session you've managed to run?**
document fetched 2026-10-10 00:21 · canonical https://reddit.com/r/ChatGPT/comments/1x1ilvp/chatgpt_pro_is_getting_insane_10_subagents_in_a