I was a top 0.01% Cursor user, then switched to Claude Code 2.0
https://blog.silennai.com/claude-code
#HackerNews #Cursor #User #Claude #Code #Switch #Coding #Journey
Cursor's latest "browser experiment" implied success without evidence
https://embedding-shapes.github.io/cursor-implied-success-without-evidence/
#HackerNews #Cursor #Browser #Experiment #Implied #Success #Evidence
1Code – Open-source Cursor-like UI for Claude Code
https://github.com/21st-dev/1code
#HackerNews #1Code #OpenSource #Cursor #UI #ClaudeCode #GitHub #DeveloperTools
The #Cursor prompt:
Remove the Bookwyrm code for now
removed nearly 300 lines of code and touched every file 😱
Removing "unnecessary" only deleted 6 lines (so the problem maybe wasn't as bad as I initially thought)
Remove all "game" related code surprisingly added on LOC, a comment.
But funny enough it made the code far clearer to understand. Where previously "gamekeys" were used now it was "order_keys". Additionally, the comment that the get_gamekeys() method returns all order keys was added (not sure if this is true)
Sidenote: Switching the variable naming to suddenly use an '_' is a bit strange.
OK let's try one more thing before wrapping up this experiment and start over with smaller steps.
Let's try to add UnitTests!
3/n
I asked #Cursor to generate unit tests for the project.
The result? Does it spark joy (or confidence)?
😬
Again, not really 😕
Nope, that is not a good unit test. This test check the setting of an internal property. This should actually be an implementation detail.
#Cursor generate roughly 1k lines of Python.
This included the (pseudo) code for querying #Bookwyrm. I would have preferred to generate this later (as a second step) But maybe it my mistake. I was prompting it wrong. I shouldn't have mentioned anything I didn't directly want.
Without looking at the Bookwyrm client code in detail, I doubt that it will work as my instance ( @realn2s@bookwyrm.social) require #2fa 🤷🏻
Checking the HumbleBundle client code I noticed several things
It contained a ton of normalization code. Of which I'm not sure if it is necessary (it might as well be, but I would leave it out for now. As it much harder to figure out that code isn't required but get executed nevertheless, than noticing that code is missing)
The authentication was done through a session cookie you had to extract form the browser. Cursor proposed to write it to an env file (that doesn't feel terrible secure 😬)
The Humble bundle client could either get the session cookie passed, or it would query the env 😬. I don't consider this good programming practice.
The code looks like it handles HumbleBundle games as well. This was requested and is additional code which shouldn't be there.
Reading through the code it looked about right. But it was too mich code, too much gold plating for me to be confident that the code really would be working.
Cursor provided no way to test the code apart from running it. And I'm not going to run it with my understanding.
On the positive side
I learned about the Click command line option "parsing" library
and about the dataclass decorator in #Python
Next step, let's try if Cursor can iteratively fixe some of my "issues"
2/n
The #Cursor prompt:
Remove the Bookwyrm code for now
removed nearly 300 lines of code and touched every file 😱
Removing "unnecessary" only deleted 6 lines (so the problem maybe wasn't as bad as I initially thought)
Remove all "game" related code surprisingly added on LOC, a comment.
But funny enough it made the code far clearer to understand. Where previously "gamekeys" were used now it was "order_keys". Additionally, the comment that the get_gamekeys() method returns all order keys was added (not sure if this is true)
Sidenote: Switching the variable naming to suddenly use an '_' is a bit strange.
OK let's try one more thing before wrapping up this experiment and start over with smaller steps.
Let's try to add UnitTests!
3/n
As my company is evaluation the use of #Cursor I'm playing around with it. First expression is mixed.
Scratching my own itch, I thought of implementing a software tool which checks which books of a #HumbleBundle book bundle I already own. Initially again my HumbleBundle library and additionally against my #BookWyrm.
I started in the planning mode.
Cursor listed some steps on how to accomplish the task and asked me which language I would like to use.
Stating that I would use whatever is best suited, Cursor picked #Python.
Being cheeky and with my preferred use in mind I asked if a browser plugin wouldn't be better suited
(imagine that, looking at a bundle, the books you own are automatically highlighted)
No chance. Cursor was determined to stick with Python!
Ok, lets generate the project
1/n
#Cursor generate roughly 1k lines of Python.
This included the (pseudo) code for querying #Bookwyrm. I would have preferred to generate this later (as a second step) But maybe it my mistake. I was prompting it wrong. I shouldn't have mentioned anything I didn't directly want.
Without looking at the Bookwyrm client code in detail, I doubt that it will work as my instance ( @realn2s@bookwyrm.social) require #2fa 🤷🏻
Checking the HumbleBundle client code I noticed several things
It contained a ton of normalization code. Of which I'm not sure if it is necessary (it might as well be, but I would leave it out for now. As it much harder to figure out that code isn't required but get executed nevertheless, than noticing that code is missing)
The authentication was done through a session cookie you had to extract form the browser. Cursor proposed to write it to an env file (that doesn't feel terrible secure 😬)
The Humble bundle client could either get the session cookie passed, or it would query the env 😬. I don't consider this good programming practice.
The code looks like it handles HumbleBundle games as well. This was requested and is additional code which shouldn't be there.
Reading through the code it looked about right. But it was too mich code, too much gold plating for me to be confident that the code really would be working.
Cursor provided no way to test the code apart from running it. And I'm not going to run it with my understanding.
On the positive side
I learned about the Click command line option "parsing" library
and about the dataclass decorator in #Python
Next step, let's try if Cursor can iteratively fixe some of my "issues"
2/n
As my company is evaluation the use of #Cursor I'm playing around with it. First expression is mixed.
Scratching my own itch, I thought of implementing a software tool which checks which books of a #HumbleBundle book bundle I already own. Initially again my HumbleBundle library and additionally against my #BookWyrm.
I started in the planning mode.
Cursor listed some steps on how to accomplish the task and asked me which language I would like to use.
Stating that I would use whatever is best suited, Cursor picked #Python.
Being cheeky and with my preferred use in mind I asked if a browser plugin wouldn't be better suited
(imagine that, looking at a bundle, the books you own are automatically highlighted)
No chance. Cursor was determined to stick with Python!
Ok, lets generate the project
1/n
Oh yeah, I use Cursor every day. I'd even say I'm a professional at it, with decades of experience.
Top-of-the-line security: 1990s. Things have changed a bit since then.
Oh no. It's true: The mouse cursor image on Wayland is delayed by one frame. 😭
Proof: https://mort.coffee/home/wayland-input-latency/
Great explaination by @lina: https://lobste.rs/s/oxtwre/hard_numbers_wayland_vs_x11_input_latency#c_edq7tn
A failed attempt to fix this in @gnome: https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4249
It sounds like there is going to be a new kernel API that would allow updating the cursor independently and therefore earlier, but I couldn't find anything about that.
Ulrike Hahn
boosted
Very thoughtful analysis by @grimalkina of the experimental design and results from the recent METR study on “the impact of early-2025 AI on experience open-source developer productivity”.
Very thoughtful analysis by @grimalkina of the experimental design and results from the recent METR study on “the impact of early-2025 AI on experience open-source developer productivity”.
David Gerard
boosted
“The users who choose Cursor are hardcore vibe addicts. They are tech incompetents who somehow BSed their way into a developer job. They cannot code without a vibe coding bot.”
I see no lie.
https://pivot-to-ai.com/2025/07/09/cursor-tries-setting-less-money-on-fire-ai-vibe-coders-outraged/
“The users who choose Cursor are hardcore vibe addicts. They are tech incompetents who somehow BSed their way into a developer job. They cannot code without a vibe coding bot.”
I see no lie.
https://pivot-to-ai.com/2025/07/09/cursor-tries-setting-less-money-on-fire-ai-vibe-coders-outraged/
As part of my job, I have to evaluate AI tools. Part of that evaluation is pushing them to their limit. Today, I realised Cursor has a setting where if you critique its work enough, it goes silent and refuses to apply changes.
It's a moody junior dev whose overconfidence and bravado quickly turn to surly silence when their work is questioned. The happy, helpful (and frequently wrong) AI is gone, replaced by a useless one with a bad attitude that won't make it past the next performance review.
Christ. I'm used to managing engineers, but I draw the line at managing AIs.