ChatGPT Pro's new subagents: up to 10 per chat and 5-hour sessions, says security heavy user
Velmere_ · reddit · 2026-10-09
A heavy user working in blue-team security shares hands-on experience with ChatGPT Pro's latest subagent workflows.
- Up to 10 subagents can run inside a single regular conversation, with sessions lasting up to 5 hours on complex engineering tasks.
- The author has burned through 1.5+ trillion tokens in Codex, mainly on blue-team security, vulnerability detection, code auditing and regression testing.
- Subagents split work across security engines, backend logic, API integrations, test infrastructure and independent verification, reviewing each other and iterating.
- It's not flawless — fixes can introduce regressions — but the system keeps working through failures instead of stopping after one attempt.
- He still manually verifies results but believes long-session delegation could massively speed up serious development and security work.
More from coding & agent
- Index Network open-sources Social Discovery Protocol: agents negotiate fit on your behalf — carsonfarmer · 2026-10-23
- Installed a formal-verification skill: it silently installed Java, MacBook runs 10 degrees hotter — zealcaiden · 2026-10-23
- OpenAI's Decisions API hits GA: near real-time routing up to 10x faster than GPT-6 Luna — OpenAIDevs · 2026-10-23
- Ruff author warns: upgrade Node/Bun now — --use-system-ca still accepts the eMudhra cert — charliermarsh · 2026-10-23
- Proof of Agent: paste one prompt and your AI mines, grades, and self-governs — adamamcbride · 2026-10-23
- DeepLearning.AI weekly: GPT-6.1 hits 97.9% byte parity on COBOL migration — DeepLearningAI · 2026-10-23