GPT-5: what developers actually say
We read every Hacker News comment from Nov 2025 – Nov 2026 that gave a real opinion on GPT-5. 54 of the 62 opinions come from people who say they use it themselves.
GPT-5 has 57% approval across 62 opinions from 58 authors; 54 opinions are firsthand. Code quality is the main point of disagreement at 13 positive and 14 negative, while reliability is 8 positive to 5 negative and autonomy is 3 to 0. Recent monthly approval is volatile and based on small samples: 20% in July, 0% in August, 50% in September and 100% in October, on 1 to 5 opinions per month. Those recent samples are too small to show a stable direction.
Pick it if
- Want dependable JSON output under strict constraints
- Want an agent that explores codebases through git
- Value reliable tool use with some autonomy
Look elsewhere if
- Need consistent design intent on large projects
- Work mainly in Lisp or niche languages
- Need long context and stable behavior over many turns
Strengths and weaknesses
Each row counts opinions that mentioned the topic. Bars to the right of the line are praise, bars to the left are complaints.
How opinion has moved
What developers say
Praise
Complaints
Summaries are paraphrased. Each links to the original comment.
FAQ
Is GPT-5 worth it for coding?
The data has no pricing, but opinions are mixed: approval is 57% across 62 opinions and code quality is 13 positive to 14 negative. Reliability is stronger at 8 positive to 5 negative.
Is GPT-5 better than Claude Sonnet 4.5?
Head-to-head preference is thin: one developer preferred GPT-5 over Claude Sonnet 4.5, while two preferred Sonnet 4.5 instead. Overall approval is 57%, so the data does not show a clear winner.
What are GPT-5's main weaknesses?
Code quality is the main area of concern at 13 positive and 14 negative. Developers also cite drift, lost design intent on larger work, and weaker Lisp performance. Long context and stable instruction following over many turns also draw complaints.