Products · Broad category
LLMs
Raised as a usability problem by 73 people on Hacker News over 4 years (2023–2026), in 77 threads.
This is a broad category rather than one product, so its issues aren't ranked on the home page.
- 1
- 1
- 1
Every issue, all time, most people first
LLM writing requires rereading to decipherUnreviewed
20232026- LLM output blathers so much that context is impossible to track2026-10-11
- LLM writing is so confusing it requires rereading and effort to decipher2026-08-11
- LLM-compressed text takes twice the effort to understand and is confusing2026-07-03
Unreviewed issues were grouped by machine: Claude Haiku labelled each comment as a usability complaint, named what it's about and summed up the problem in a line, and a small model grouped similar problems. Nobody has checked them by hand. Expect mistakes.
The newest 50 complaints
- UI is mushed together on Android mobile browsers, forcing desktop site mode2026-10-19
- LLM output blathers so much that context is impossible to track2026-10-11
- Results were so poor the model was borderline unusable2026-10-02
- Typing a full sentence for every action makes repeated tasks unusably slow for proficient users2026-09-26
- LLMs report the first Google results, needing much more prompting for niche answers2026-09-18
- Autoplay on load makes it hard to step through tokens one at a time2026-09-12
- Pause resets the animation to its start instead of pausing, so highlighting disappears2026-09-09
- LLM writing is so confusing it requires rereading and effort to decipher2026-08-11
- Floods analysts with false positives, so real alerts get ignored2026-07-22
- Reviewing and refining agent output requires copy-pasting snippets and describing issues one message at a time2026-07-13
- LLM-compressed text takes twice the effort to understand and is confusing2026-07-03
- Too many worktrees make it hard to find the actual file to edit2026-07-03
- Typing lags for seconds after a few conversation turns on iOS Safari2026-06-10
- Getting tool calls to work with local models feels counterintuitive2026-04-14
- Conversational text interface slows interaction, forcing users to dig through prose for key information2026-03-03
- Waits about 1.5 minutes before the first token, then slow output, making it practically unusable2026-03-01
- Completed uploads are unusable until you poll for a separate processing step2026-02-08
- Can't steer the model; it repeats errors even after repeated explanations2026-02-04
- Learning to use LLMs effectively for coding is unintuitive and difficult2026-01-27
- Takes dozens of retries and pasted-back errors to get valid query syntax2026-01-12
- Interfaces feel unpredictable and unreliable, so users can never quite trust them2025-12-19
- Juniors constantly fight unhelpful autocomplete suggestions and sometimes accept them and get lost2025-12-07
- Hard to figure out where and how to configure things2025-11-29
- Gives confident fixes that turn out broken, so non-experts can't tell good output from bad2025-10-27
- Getting the last 20% of classification accuracy takes endless prompting and tuning iterations2025-10-21
- Constant scrolling is distracting, so the reader couldn't follow the article2025-09-06
- Tools are unintuitive and pretend to be easy, so people must experiment heavily to learn them2025-09-05
- Too many AI tools; finding the one that fits is time-consuming2025-08-05
- Hard to find a newly announced model's cost across many marketing pages2025-07-25
- Repeats the same mistakes and delivers poor work that takes longer to fix than doing it yourself2025-07-19
- Generated code often fails, and debugging it takes longer than rewriting2025-07-15
- Different LLM components expect different schemas and chat templates, making the flow confusing2025-07-13
- Splitting content into many small pages is frustrating, and mobile hides the table of contents2025-07-11
- Edits sometimes applied incorrectly, forcing developer to hunt for the bug2025-07-07
- Website is horrible and impossible to navigate2025-06-30
- LLMs are nearly impossible to use right, requiring real skill2025-06-03
- Tools have a steep learning curve to get good results2025-05-29
- Hard to migrate workflows between LLMs because features and behaviors differ2025-05-22
- AI-generated email drafts get in the way and must be tweaked2025-05-20
- Going back to retry a prompt and reverse traversal is manual and tedious2025-05-16
- LLMs hallucinate polars functions as pandas functions, forcing documentation checks2025-05-05
- Hard to use because results are unreliable and you must learn workarounds2025-04-12
- Can't tell whether it did what was expected; checking takes lots of work2025-03-28
- Forced to format model inputs as separate conversation messages, with no plain completion option2025-03-18
- Navigation seems confusing and the UX is too crowded on screen2025-01-28
- Chat interface feels clunky to use when programming with LLMs2025-01-07
- Filters cannot be turned off, blocking harmless tasks like translating movie subtitles2025-01-03
- Hard to navigate text because it constantly changes while switching focus2024-12-11
- Having two different search patterns to remember is annoying and hard to adopt2024-11-13
- Hard to tell which tasks LLMs can reliably solve2024-09-26