01 · The big storyClaude Code
−14 pts
Claude Code's approval has fallen 14 points in a year.
It is still the most discussed tool on HN, but reliability complaints now outnumber praise 296 to 107.
Read the analysis → Nov 2026 edition / Data through Nov 2, 2026
The AI coding tools developers actually trust.
We read 20,064 Hacker News comments and extracted 13,197 opinions to rank 22 tools and 45 models. No benchmarks. No sponsors. Just what working developers said.
Independent20,064 comments screened13,197 opinions extracted
The dataset
Public Hacker News comments, Nov 2025 – Nov 2026.
20,064comments screened
13,197opinions extracted
22tools ranked
45models ranked
The ranking · Nov 2026
Coding tools, ranked by developer approval.
Approval is the share of opinions that are positive, adjusted for sample size. Every score links back to the comments behind it.
| # | Tool | Type | Approval | Last 90 days | Opinions |
|---|---|---|---|---|---|
| 01 | PiOpen source | Terminal agent | 84%±2.9 | 80% | 423 |
| 02 | OpenAI CodexOpenAI | Terminal agent | 69%±1.7 | 63% | 2,013 |
| 03 | OpenCodeOpen source | Terminal agent | 66%±2.8 | 55% | 780 |
| 04 | ZedZed Industries | Editor | 59%±3.8 | 57% | 457 |
| 05 | Claude CodeAnthropic | Terminal agent | 54%±1.5 | 51% | 3,026 |
| 06 | CursorAnysphere | AI IDE | 55%±3 | 53% | 766 |
| 07 | AntigravityGoogle | AI IDE | 45%±4.7 | 42% | 301 |
| 08 | GitHub CopilotGitHub | IDE extension | 42%±3.2 | 38% | 659 |
| 09 | Gemini CLIGoogle | Terminal agent | 29%±5.1 | 25% | 211 |
0184%0269%0366%0459%0554%0655%0745%0842%0929%
Pi
Open source
OpenAI Codex
OpenAI
OpenCode
Open source
Zed
Zed Industries
Claude Code
Anthropic
Cursor
Anysphere
Antigravity
Google
GitHub Copilot
GitHub
Gemini CLI
Google
Rank uses the lower bound of a 90% confidence interval, so a 55% score on 766 opinions can sit below a 54% score on 3,026. Shown: tools with 100+ opinions.
All 22 tools →What changed
Three things worth knowing this month.
02 · The switch warsOpenAI Codex
181–62
Codex is winning the switch wars.
When developers named a preference between Codex and Claude Code, they picked Codex 181 times to 62.
See the comparison →03 · The quiet favouritePi
84%
Pi is the quiet favourite.
The open-source terminal agent tops approval on 423 opinions — with a customization score of +146.
See what developers say →Head-to-head
Comparisons people actually make.
VS
Claude Code vs Cursor
Developers pick Claude Code 61–42, but approval is a dead heat.
54%103 direct picks55%
Read the comparison →VS
Claude Code vs OpenAI Codex
OpenAI Codex wins the direct preference 181–62. Claude Code still wins on autonomy.
54%243 direct picks69%
Read the comparison →VS
OpenAI Codex vs Cursor
OpenAI Codex is the pick over Cursor 21–7, but Cursor has more total opinions.
69%28 direct picks55%
Read the comparison → The Shift Report
One email a month. The tools gaining and losing developer trust.
The five-minute read that turns 20,000 Hacker News comments into a clear picture: the biggest swings, the complaint everyone repeats, and who switched — with links to the original comments.
- The month's biggest approval swings, with sample sizes
- One deep dive, one chart, three quotes worth reading
- No sponsors. No hype. Unsubscribe in one click.
The next editionMonthly · 5 min
What lands in your inbox
The tools that moved, the comments that explain why, and the comparisons developers are actually making.
Read the latest issue →Methodology
How the ranking is built.
We search Hacker News comments month by month, keep the ones that contain a real opinion about a tool, and tag each one for sentiment, first-hand use, topics and switches. Approval is the positive share of opinions that took a side; rank uses the lower bound of a 90% confidence interval.
Sources: public comments on Hacker News, collected through the HN Search API by Algolia, Nov 2025 – Nov 2026. Every paraphrased opinion on tool pages links to its original comment. No vendor can pay for placement.