Show HN: State of the Art of Coding Models, According to Hacker News Commenters

Coders crown Claude, then drag it for outages, prices, and mood swings

TLDR: A daily tracker is turning Hacker News comments into a scoreboard for coding AI tools, showing which names people talk about most and how they feel about them. The big debate: popular tools like Claude get tons of attention but also tons of complaints, while open and local alternatives are gaining fans fast.

A new Hacker News-powered tracker is trying to answer the internet’s favorite nerd argument: which coding AI is actually winning right now? The project scans popular posts, has an AI sort through comments, and logs which coding tools people mention most — plus whether those mentions are glowing, grumpy, or somewhere in between. In pure popularity, Claude is the chatter king. But in the comments, that crown comes with a side of tomatoes.

The community reaction is the real soap opera. One of the loudest themes is that people like Claude’s coding help, but hate the baggage: pricing complaints, server downtime, and the feeling that the service can go from genius to useless depending on the hour. Meanwhile, GPT-5.5 is getting quieter but warmer praise, which sparked a classic comments-section split: is the most-mentioned tool really the best, or just the one people complain about most? Others are practically starting a rebellion, saying the future belongs to local and open models because they’re tired of what one commenter described as wildly unpredictable performance from the big-name systems.

And yes, there was drama. Open models like kimi, qwen, and deepseek got a surprising amount of love, which led to a spicy little anti-corporate victory lap. Another commenter basically barged in with, “Forget models — where’s the ranking for coding agents?” In other words: the leaderboard is live, the nerds are fighting, and the comments are treating AI coding like a reality show finale.

Key Points

  • The article describes a daily updated project that tracks coding model popularity and sentiment from Hacker News comments.
  • Its pipeline collects the top 200 Hacker News posts from a 24-hour period and uses an LLM to select up to 50 posts about LLMs or coding.
  • Selected post titles and comments are sent to Gemini, which identifies models from the OpenRouter model list and rates sentiment per comment.
  • The project logs comment IDs and sentiment outputs to Google Sheets to make the process auditable and easier to debug.
  • The embedded chart summarizes Top 10 model popularity using a 10-day trailing aggregate covering 2026/4/22 to 2026/5/1.

Hottest takes

"Claude is currently taking the #1 spot in mentions, it carries a lot of negative sentiment due to API pricing policies and frequent server downtime" — jdw64
"wildly unpredictable but systematic performance... at different times of day or week they are getting degraded beyond belief" — dzink
"the real salvation will be a combination of more mature OS models" — fredcallagan
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.