"Read" tool should contain line
Resolved 💬 4 comments Opened Feb 26, 2026 by jenglong1899 Closed Apr 3, 2026
💡 Likely answer: A maintainer (guidedways, contributor)
responded on this thread — see the highlighted reply below.
What variant of Codex are you using?
CLI
What feature would you like to see?
If the AI only reads 1 to 200 lines, I hope I can be informed, so that perhaps it hasn't read enough code, and I can notice it in time.
Additional information
_No response_
4 Comments
@Linde7777 there's no need to do this with codex. Codex is a frontier SOTA model that know what to read and what not to read. You should describe the issue in your prompt and be descriptive where there's ambiguity and let codex manage the context on its own. There's never going to be a situation where it will read 1-200 lines and miss out on additional context simply because it didn't read enough. It knows how many to read given the problem at hand.
@Linde7777, can you describe the problem you want to solve here? You've proposed a solution, but I'm not sure what issue you're trying to address.
If line numbers could be displayed, it would give me a greater sense of control, since LLM sometimes make mistake, the sooner I found it, the faster I can correct.
Additionally, the issue might be with the read tool itself—the read limit is really too small. There's a memory document in my codebase, and it's actually not very large. At least a week or two ago, every time it read this file, the context would only drop from 100% to 99%. But in the last two days, I found it have to read it multiple times, after reading it, the context is left at only 97% or 95%.
I'm using GPT 5.2 high.
I also feedback in cli, the thread id is:
019c9dc1-03e1-72f3-b42e-3ceb1ffb7738
<img width="1920" height="720" alt="Image" src="https://github.com/user-attachments/assets/3516b2bd-ac98-431b-9d7c-1af8b709d152" />
This feature request hasn't received enough upvotes, so closing.