Products · Product
Claude Code
Raised as a usability problem by 306 people on Hacker News over 2 years (2025–2026), in 233 threads.
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
Every issue, all time, most people first
Terminal redraws everything repeatedly, causing scroll bugs on long outputUnreviewed
20252026- Terminal interface redraws the whole screen on every keystroke, which is maddening to watch2026-09-15
- Typed characters sometimes not visible in the terminal UI, plus rendering glitches2026-07-19
- Duplicate and offset lines appear, scrollback loses lines, and the terminal shows garbage2026-06-23
Asks permission repeatedly for routine actions, pushing users to skip permissionsUnreviewed
20252026- Human approval step fails once the system becomes reliable, so people stop checking2026-08-11
- Asks permission every time for basic safe commands instead of trusting a tool once2026-08-10
- Approving each agent command manually is impractical, creating too many prompts to review2026-08-08
Plan mode gives plans random file names and struggles with multiple plansUnreviewed
20252026- Plan mode jumps to huge, off-base plans, forcing wasted turns fixing intent2026-10-04
- Plan mode no longer lets users iterate and explore options before work starts2026-10-03
- Plan Mode doesn't stop it from running bash and changing files2026-09-14
Breaks standard copy and paste behaviour on LinuxUnreviewed
20252026- Copy-paste was randomly changed, making it annoying to copy text2026-07-17
- Copy/paste behaves differently than users expect, breaking a basic OS-wide pattern2026-07-17
- Copy/paste behaves differently from other command-line apps, which is unexpected2026-07-17
Agent repeatedly fails basic tool use, forcing user interventionUnreviewed
20252026- Sub-agents are opaque; user can't follow what they do or steer them2026-09-30
- Agent tool calls fail with permission errors and must-read-before-write blocks, wasting tokens2026-08-20
- Switching between many agents gets confusing; user forgets which agent is stuck or working2026-08-09
First typed character lands in wrong spot and can't be backspacedUnreviewed
20252026- Cursor doesn't advance after typing a space, so the next character position is unclear2026-09-28
- First typed character lands in the wrong spot and can't be backspaced2026-07-19
- Holding backspace moves the cursor instead of deleting characters, making prompt editing painful2026-01-08
Stops work midway, so user must repeatedly prompt it to continueUnreviewed
20252026- Claude stops doing work, so user must repeatedly prompt it to continue2026-04-23
- Can't interrupt Claude, UI freezes, sessions disconnect and take long to reappear2026-02-25
- Claude hands control back to the user repeatedly without accomplishing the task2025-09-29
Users can't see which files the agent read or searchedUnreviewed
20252026- Can't see whether the agent is deleting files or running commands behind hidden output2026-07-20
- Users can't see what files are being edited or what subagents are doing2026-03-30
- Hidden file names make the agent's actions a black box, so the user can't catch mistakes2026-02-16
OAuth assumed to work with no docs saying a separate API key is neededUnreviewed
20252026- Assumed OAuth would work; no docs pointed to needing a separate API key2026-04-06
- Getting the OAuth token into another user's keychain took lots of fiddling2025-12-15
- Forced to sign into a user account and create an API account instead of entering own key2025-03-09
Interface sluggish to respond to typed inputUnreviewed
20252026- Interface feels subpar, and the tool is slow and annoying to use2026-04-15
- Input often not registered, and the interface feels slow to respond2026-03-24
- Interface is slower than desired, sluggish to respond to input2025-10-28
Spends hours being steered to correct output, with context lostUnreviewed
20252026- Context silently dropped during long sessions, user can't find or see the change2026-04-24
- Unavoidable pauses while it summarises its context window, shown constantly2026-02-06
- Spends hours being wrangled into correct output, with context lost sporadically2025-06-16
Can't run scripts on phone, so hard to verify what Claude didUnreviewed
20252026- Hard to see what Claude actually did; ctrl+o and ctrl+e hide details and verbose flags show everything2026-03-02
- Can't run scripts on phone, so it's hard to verify what Claude did2026-01-04
- Claude Code almost completely unable to build menu screens and UI elements2025-07-27
Agent ignores instructions and planning docs and won't stop when toldUnreviewed
20252026- Agent says it will start, then sits waiting for user input; steering mistakes2026-07-27
- Agent ignores instructions and planning docs, and won't stop when told to2026-06-22
- Skills don't trigger reliably and the agent doesn't follow instructions to change modes2026-02-24
Typing a prompt accidentally answers Y/N questionsUnreviewed
20252026- Typing a prompt accidentally answers Y/N questions, steering or stopping the agent2026-10-31
- Permission prompts appear mid-typing and swallow keystrokes, answering unintentionally2026-05-05
- Random misbehavior that prompting and rules can't prevent; manual fix would be quicker2026-04-07
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
- Feedback suggestions appear in session even though they cannot be submitted2026-11-03
- Feels bloated and painfully slow to use2026-11-01
- Developers stuck in a text box with a diff viewer, unable to see project's logical pieces2026-11-01
- Typing a prompt accidentally answers Y/N questions, steering or stopping the agent2026-10-31
- Dense text wall makes it hard to search, select, or split related content2026-10-31
- Descriptions are confusing, so reader can't tell which version applies2026-10-30
- Feedback popup appears mid-session and requires a keypress to dismiss, interrupting work2026-10-27
- Changes affect workflows by default, without notice, documentation, or a way to turn them off2026-10-16
- Rewind wipes filesystem changes by default, a poorly documented behavior people must work around2026-10-14
- Switching away from an ongoing conversation loses it2026-10-14
- Remote-control server turned on by default silently, with no explanation or notification2026-10-13
- Rewind is no longer accessible from the UI; can't rewind to steering messages2026-10-12
- Accidentally submits ratings when the user meant to type other input2026-10-07
- Settings keep changing and re-enabling data collection the user turned off2026-10-07
- Adding or editing a suggested response takes many steps2026-10-07
- Must manually copy conversation context to another account when usage limit is hit2026-10-06
- Plan mode jumps to huge, off-base plans, forcing wasted turns fixing intent2026-10-04
- Plan mode no longer lets users iterate and explore options before work starts2026-10-03
- Sub-agents are opaque; user can't follow what they do or steer them2026-09-30
- Cursor doesn't advance after typing a space, so the next character position is unclear2026-09-28
- Too many open tabs make it hard to locate old AI sessions2026-09-25
- Model keeps running arbitrary Bash commands instead of built-in tools, making hooks and permissions unusable2026-09-23
- Long Claude prose reports are hard to decipher, so the user wants choice buttons instead2026-09-22
- Kicked back to login weekly, and loses open window state after signing back in2026-09-22
- Terminal sits in the sidebar, and dragging it to the bottom or toggling it resets it2026-09-21
- Waiting for tasks to finish gives poor progress feedback compared to Codex2026-09-20
- Existing harness doesn't give easy visibility or good recap of work2026-09-15
- Terminal interface redraws the whole screen on every keystroke, which is maddening to watch2026-09-15
- Using AI coding agents for art creates a lot of friction and paper cuts2026-09-14
- Plan Mode doesn't stop it from running bash and changing files2026-09-14
- The user interface is a little confusing to use2026-09-11
- Keeps re-asking to enable data sharing after the user declined, disguised as a task approval2026-09-10
- Commenter struggles to get Claude to use LSP at all2026-09-04
- Default keybindings clash in most terminals, so keys trigger the wrong commands2026-08-29
- Takes too long and runs too many unwanted steps before writing code2026-08-26
- Screen turns into two columns of interleaved garbage when the phone is rotated2026-08-25
- Can't preview pasted images and can't edit mid-sentence text like a normal interface2026-08-22
- Agent tool calls fail with permission errors and must-read-before-write blocks, wasting tokens2026-08-20
- Silently forces bash over dedicated tools, ignoring LSP and prompts, with low adherence to overrides2026-08-20
- Non-technical coworkers can't run command-line tools or open .md files, so they need GUIs.2026-08-12
- Human approval step fails once the system becomes reliable, so people stop checking2026-08-11
- Asks permission every time for basic safe commands instead of trusting a tool once2026-08-10
- Unix commands with mysterious flags can't be whitelisted, forcing hard-to-review manual permission approvals2026-08-10
- Selecting chat option makes it think and ask what to discuss, wasting a turn2026-08-10
- Output viewer is broken and shows nothing, as the model redirects output to scratch files2026-08-10
- Commands Claude wants to run are truncated in terminal, so user can't review them2026-08-10
- Switching between many agents gets confusing; user forgets which agent is stuck or working2026-08-09
- Approving each agent command manually is impractical, creating too many prompts to review2026-08-08
- Setting up messaging between sessions requires manually opening terminals and copy-pasting messages, feels clunky2026-08-08
- Non-programmers struggle to install it and to use the shell interface2026-08-04