GPT-5

High reasoning effort on GPT-5 (GPT-5.1) is worth the wait for hard bugs. For a gnarly concurrency bug, cranking reasoning effort up took longer but it actually found the root cause instead of a band-aid fix.

Reply

JA

Jamie L.

1mo ago

Reproducible bug: Cursor mangles multi-file edits on GPT-5 (GPT-5.1). When a single instruction touches 3+ files, about 1 in 5 times one of the edits gets applied to the wrong file. Easy to catch in review but still concerning for anyone auto-applying diffs.

5

SA

Sam H.

1mo ago

Inconsistent output for the identical prompt on GPT-5 (GPT-5.1). Ran the exact same prompt twice in a fresh session and got meaningfully different code structure both times — expected some variance, not this much.

3

BI

Bill Ferreira

1mo ago

Reasoning quality on GPT-5 (GPT-5.1) varies a lot by task type. Excellent on algorithmic problems, noticeably weaker on ambiguous product-requirement style questions where it needs to make a judgment call.

5

RE

Ren T.

16d ago

Went back to the previous version of GPT-5 (GPT-5.1) after trying the new one. The new version is smarter in isolated benchmarks but in my actual workflow it second-guesses itself more and takes longer to converge on a fix.

BI

Bill Ferreira

18d ago

Quality feels inconsistent session to session on GPT-5 (GPT-5.1). Some sessions it's excellent, others it makes basic mistakes on things it clearly should know. Hard to predict which you'll get.

3

J.

J. Alvarez

5d ago

GPT-5 (GPT-5.1) is genuinely great at refactoring large files now. Asked it to split a 2,000 line controller into smaller services and it did it cleanly, preserved behavior, and even flagged an existing bug it found along the way.

2

MA

Marcus D.

12d ago

Overall quality bump with GPT-5 (GPT-5.1) is real. Not hype — day to day the suggestions feel more aligned with what I actually meant, fewer 'technically correct but useless' answers.

EL

Elina R.

21d ago

Formatting bug in GitHub Copilot's diff output. Trailing whitespace keeps getting added to every line it touches, which is making every PR noisy with whitespace-only diffs on unrelated lines.

OM

Omar S.

6d ago

Cursor Pro limits: fine for solo work, tight for pairing. For my day-to-day solo coding Cursor Pro is plenty. The moment I try to use it for a full pairing session with a teammate watching, we blow through it in under two hours.

AI

Aino M.

28d ago

GitHub Copilot uptime has been rock solid for me this month. After a rocky start a couple months back, I haven't had a single failed request in weeks. Whatever they fixed on the backend, it's working.

JA

Jamie L.

26d ago

Cursor Pro price increase discussion. Not thrilled about the price change, especially since the usage caps didn't move proportionally. Curious how others are reacting.

3

SA

Sam H.

8d ago

GitHub Copilot crashed mid-task twice today. No error message, just the CLI process dying silently. Had to restart and lost the in-progress edit. Happened right after the latest update.

2

JA

Jamie L.

1mo ago

GitHub Copilot's tool-calling for running tests is solid. It runs the test suite after changes, reads the failures, and iterates without me having to copy-paste output back in. That loop alone is worth the subscription.