GPT-5.6: what 33 developers actually say.
33 Hacker News opinions from 34 developers, Nov 2025 – Nov 2026. 26 come from people who say they use it. Here is what they praise, what they complain about, and who switched.
The case for GPT-5.6, in one paragraph.
GPT-5.6 earns 76% approval across 33 opinions. Developers are divided on it. In the last 90 days it sits at 75%.
Pick it if you
- Want root-cause fixes in an existing codebase
- Care about clean interfaces and low token use
- Prefer clear communication from the agent
Look elsewhere if you
- Need strict instruction following on scoped tasks
- Need an agent that never refuses security work
- Are on a tight budget
Where it wins and loses.
Mentions across 33 opinions. Bars are counts, not votes; a tool can be both praised and criticized on the same topic.
View data table
| Topic | Praise | Complaints | Mentions |
|---|---|---|---|
| Code quality | 11 | 2 | 13 |
| Reliability | 5 | 5 | 10 |
How opinion has moved.
Monthly approval, Jul 2026 – Sep 2026. Bars show opinion volume; low-volume months are still included but faded.
View data table
| Month | Approval | Opinions |
|---|---|---|
| Jul 2026 | 73% | 13 |
| Aug 2026 | 89% | 9 |
| Sep 2026 | 56% | 8 |
What developers say.
Summaries are paraphrased. Each links to the original Hacker News comment.
Praise
4 selectedGPT-5.6 correctly identifies root causes of problems in their codebase.
Models since GPT-5.6 understand high-level intent well enough that they stopped writing spec rubrics.
GPT-5.6 was and still is much better, after criticizing the Astra and Sol models.
Recent GPT-5.6 experiences were very positive, without the erratic behavior they saw elsewhere.
Complaints
4 selectedGPT-5.6 became unusably bad at following instructions in their coding agent before the GPT-6.0 release.
GPT-5.6 is smarter than DeepSeek but too pricey, and sometimes refuses or stops tasks midway.
GPT-5.6 refused to follow security-testing prompts and even declined to edit skill files.
GPT-5.6 in Codex is unusable: ignores scoped tasks, deletes unrelated code, and adds regressions.
Questions developers ask.
Is GPT-5.6 worth it?
Is GPT-5.6 better than Claude Opus 4.8?
How reliable is GPT-5.6 for coding?
The data behind this page.
Developers give GPT-5.6 a 76% approval score across 33 opinions, and 26 of those come from firsthand use. Code quality is the strongest signal at 11 positive and 2 negative mentions. Reliability is split at 5 positive and 5 negative, with complaints about ignored scope, deleted code, and refused tasks. Monthly approval fell from 89 in August to 56 in September on 8 opinions, so the 90-day score of 75 hides a recent dip.
Every number comes from public Hacker News comments collected through the HN Search API. A language model tagged each comment for sentiment, first-hand use, topic and switches; a human spot-checks samples. No vendor can pay for placement.
Get the monthly report
Which AI coding tools gained or lost developer approval on Hacker News, and why people switched. One email a month.