Data and AI struggles

Top rising problems today

What people in data, tech and AI struggle with, from public posts: 1,143 problems so far from 6,620 posts, +814 posts this week across 286 companies and products. Scanned daily.

Problem posts per day

Problem posts per day5,100 posts, Sep 22 to Oct 21

Rising today

All rising
LLM appsSteady38d active+19 this weekNo reliable way to evaluate LLM app quality103 posts · 98 people · on GitHub issues, Hacker News, RedditOutages and changeRising23d active+10 this weekClaude Code upgrades break working setupsClaude Code · 37 posts · 32 people · on GitHub issues, Reddit, Hacker NewsOutages and changeRising10d active+6 this weekGitHub Copilot upgrades break working setupsGitHub Copilot · 14 posts · 13 people · on GitHub issuesDev toolingRising4d active+5 this weekGitHub Copilot feels slow and laggyGitHub Copilot · 7 posts · 7 people · on Reddit, GitHub issues, Hacker NewsMLOpsRising17d active+7 this weekRunning llama.cpp locally hits memory and speed limitsllama.cpp · 29 posts · 26 people · on GitHub issues, RedditLLM appsRising12d active+5 this weekModels fail to return valid structured output17 posts · 15 people · on GitHub issues, Hacker News, RedditCareersRising18d active+10 this weekTech job hunting is brutal right now32 posts · 30 people · on Reddit, GitHub issues, Hacker NewsAI codingRising6d active+5 this weekTrouble with GitHub Copilot harnessGitHub Copilot · 8 posts · 8 people · on GitHub issues, RedditMLOpsRising5d active+4 this weekNVIDIA GPU access and drivers are painfulNVIDIA · 7 posts · 7 people · on GitHub issues, MastodonOutages and changeRising12d active+4 this weekvLLM upgrades break working setupsvLLM · 14 posts · 14 people · on GitHub issuesSecuritySteady29d active+5 this weekVulnerability scanners flood teams with noise262 posts · 262 people · on fru.dev sites, Reddit, GitHub issuesMLOpsRising4d active+3 this weekFine-tuning Hugging Face is costly and unpredictableHugging Face · 5 posts · 5 people · on GitHub issuesOutages and changeSteady12d active+5 this weekOllama upgrades break working setupsOllama · 15 posts · 11 people · on GitHub issuesAI codingRising10d active+5 this weekTrouble with Cursor agentCursor · 13 posts · 13 people · on Community forumsDev toolingRising3d active+3 this weekClaude Code error messages are crypticClaude Code · 4 posts · 4 people · on GitHub issues

Persistent problems

All persistent

Questions

What is a problem here?

A pain people state in public: a question nobody has answered well, a complaint, a workaround, a bug with many reactions. Posts that say the same pain about the same company or product are grouped into one problem with a short plain statement.

Where do the posts come from?

Public sources only: Hacker News (Algolia API), public Reddit feeds, Stack Overflow, the most-reacted GitHub issues of major data and AI repos, Discourse community forums (dbt, OpenAI, Hugging Face, PyTorch and more), Bluesky and Mastodon public posts, YouTube video titles, 1 and 2 star App Store reviews, and outages, deprecations, breaches and advisories from other fru.dev sites. Slack is read only where a community publishes public archives. No logins, nothing behind a sign-in.

How are posts grouped?

By rules, so every grouping can be explained: a post counts when it states a pain ("struggling with", "doesn't work", "workaround", "is there a way"), it is matched to a known issue (rate limits, schema drift, warehouse cost, flaky CI and so on) and to the companies and products it names. Posts about a product with no known issue are grouped by the words they share. The Method page lists every rule.

What do rising, persistent and new mean?

Rising: more posts in the last 7 days than in the 7 before. Persistent: raised on 5 or more days across at least two weeks, and still active this week. New: first posted in the last week. Problems backed by fewer than 3 posts from 2 people or places are marked unverified.

How can I use this?

As a map of unsolved pain: common, persisting problems are the signal to build something. Every problem links to the posts behind it, the companies involved and its trend over time. The data is also open as JSON, llms.txt and an OpenAPI spec for agents.