Up 9 points: Pi.
71 opinions in the latest month. The all-time score is 84%.
Read the evidence →The tools gaining and losing developer trust, the complaint everyone repeats, and who switched. Built from 20,064 public Hacker News comments, Nov 2025 – Nov 2026.
71 opinions in the latest month. The all-time score is 84%.
Read the evidence →177 opinions in the latest month. The all-time score is 54%.
Read the evidence →93 opinions in the latest month. The all-time score is 66%.
Read the evidence →Claude Code still has the largest opinion base on Hacker News. Its approval has moved from 68% at the start of the window to 54% all-time, and 51% over the last 90 days.
| Month | Approval | Opinions |
|---|---|---|
| Nov 2025 | 68% | 174 |
| Dec 2025 | 71% | 251 |
| Jan 2026 | 58% | 365 |
| Feb 2026 | 55% | 347 |
| Mar 2026 | 58% | 290 |
| Apr 2026 | 42% | 383 |
| May 2026 | 49% | 253 |
| Jun 2026 | 54% | 216 |
| Jul 2026 | 50% | 189 |
| Aug 2026 | 48% | 203 |
| Sep 2026 | 49% | 166 |
| Oct 2026 | 56% | 177 |
Claude Code over 10 months of full-time work gets better at building the right thing, though it needs prompting for refactors.
Claude Code nearly one-shotted most game stages from voice-transcribed prompts.
Claude Code is bloated and slow and gives no better results than other harnesses.
Hitting Claude Code usage limits a few times blocked even simple questions in chat.
Complaints outnumber praise 296 to 107 on reliability. Watch whether a release changes that ratio.
181 developers picked Codex over Claude Code, against 62 the other way.
Pi and OpenCode still score well on smaller samples. More opinions are the test.
The next edition lands at the start of the month. Corrections and data changes are logged in the changelog.
Which AI coding tools gained or lost developer approval on Hacker News, and why people switched. One email a month.