You can turn ChatGPT into Claude: an employee Tibo showed how to triple the context window. That lets the model retain much more of the conversation history and pick things up from just a few words:
There is a downside: it will significantly increase token usage. But in many cases, it will save a lot of time and frustration.
@typespace_ai
Open ~/.codex/config.toml and add or update these settings at the top level, before any [section] headers:
```
model = "gpt-5.6-sol"
model_context_window = 1000000
model_auto_compact_token_limit = 900000
```
There is a downside: it will significantly increase token usage. But in many cases, it will save a lot of time and frustration.
@typespace_ai