Exactly how I feel. If I wanted this agent-centric view without being able to easily see the code I would be using Claude Code.
I use Cursor because agents are not ready to be the ones driving. I need to drive. I still need to understand all the code (and easily browse it) and keep a close watch over what the AI is doing.
I did some experimenting recently and I'm quite convinced that when I use Comcasts DNS they are selling it to advertisers. I've switched to 1.1.1.1 simply because it annoys me that Comcast is doing this.
> So, try brainstorming the issue with Claude Code, talk it through so it's on the same page as you, ensure it's done research (web search, docs) to weigh the best solutions, and then enter plan mode so it generates a markdown plan file.
From there you can read/review,tweak the plan file. Or have it implement it. Or you implement it.
This is exactly the workflow that works very well for me in Cursor (although I don't use their Plan Mode - I do my version of it). If you know the codebase well this can increase your speed/productivity quite a bit. Not trying to convince naysayers of this, their minds are already made up. Just wanted to chime in that this workflow does actually work very well (been using it for over 6 months).
They said exactly that they do not think these are double bad.
They are presenting them as examples of things that a lot of people do say, and many laws are written this way, and many cops, prosecutors, & judges treat them as double bad.
Speeding without a seatbelt is two separate infractions so it should be double bad. Just like robbing a bank and shooting someone is double bad compared to either individually.
And intentional killing is generally considered worse because it means you thought about it and then did it, vs when it's due to acute emotional disorder. Intentional crimes are usually treated more harshly
I'm not talking about two crimes together... I'm talking specifically about changing the punishment based on intent or secondary effects.
Owning a gun is legal, but if you've ever smoked weed it's illegal for you... murder is bad, but if it's because you hate a protected group it's a much harsher punishment... in some places, you can't be pulled over for not wearing a seatbelt, but you can be charged with it if you're pulled over for anything else. It's just to attach additional charges for prosecution.
I don't like the idea of excessive charges as part of the prosecution process in general. It creates/extends what I consider an unfair asymmetry between the state (prosecution) and the individual. That's not to say there aren't similar examples in the other direction, such as a clerical error resulting in dismissed charges altogether.
Are their hooves, though? The fossil record clearly shows a progression in their ancestors from having feet with many toes to the single "toe" they have now.
How is this a bad thing for them? Now they don't have to pay for clicks from people who aren't interested in their services. People who want to hire a lawyer will still click.
The idea for all this content marketing spam is that writing those articles (eg. "how can I dispute a speeding ticket?") is cheaper than buying ads for "speeding ticket lawyers" or whatever. If people don't click on those articles, the whole strategy falls apart.
I have been noticing this myself for the last couple of months. I cannot get the agent to stop masking failures (ex: swallowing exceptions) and to fail loudly.
That said, the premise that AI-assisted coding got worse in 2025 feels off to me. I saw big improvements in the tooling last year.
I keep finding myself saying “stop over complicating things” over and over again, because even the simplest questions about how to load a file sometimes gets a code response that’s the size of a framework.
Good news. Been using Cursor heavily for over a year now (on the Ultra plan currently). Hope we get access to this as part of our existing subscriptions.
I use Cursor because agents are not ready to be the ones driving. I need to drive. I still need to understand all the code (and easily browse it) and keep a close watch over what the AI is doing.
reply