I disagree with Dan Luu’s style choices as well but the style isn’t due to lack of ambition or effort. He has put a lot of thought and measurement into it:

Another funny thing here is that the people who appreciate the styling generally appreciate that the site doesn’t override any kind of default styling, letting you make the width exactly what you want (by setting your window size how you want it) and it also doesn’t override any kind of default styling you apply to sites.

https://danluu.com/slow-device/

score 1 · canonical

Thanks, I appreciate these changes and the explanation behind them.

I respect people’s desire not to avoid posts that are largely LLM-generated or human-written posts about work that is mainly LLM-generated, but I think “vibecoding” as a tag was too broad, and Fillipo’s post was a good example of the tag being used in a “spoonful of sewage” way that always felt more like a witch hunt than a good faith way to help people filter out largely LLM-generated work.

score 1 · canonical

Enough, Simon. You have carried enough water for these companies.

I found this personal attack distasteful and unnecessary.

score 1 · canonical

The first quote follows the sentence “when I first started using my coding agent,” which was two years ago. It’s not describing the situation today.

The second quote is accurate today of Claude Code and Codex. They expect access to everything by default and it’s the user’s responsibility to figure out how to limit access. Even with Claude Code’s sandbox, I don’t see a way from the docs to restrict read access to a single repo directory.

score 1 · canonical

Which part is wrong?

Sandbox is off by default in Claude Code. Even when you enable Claude’s sandbox, my understanding is that it still gives itself read access to your whole system, and you have to enumerate every path it shouldn’t be able to read. I’m reading the docs now and having trouble understanding how I’d even configure it to say, “You can only read the current directory.”

My point is that Claude Code and Codex are insecure by default, and they make you do a lot of work to restrict their access rather than designing the agents to use the least privileges necessary to do their work.

score 1 · canonical

Thanks for reading!

Which wish list items do you mean? The models are capable of these things but the agents don’t take advantage.

score 1 · canonical

Thanks for reading!

For example, I don’t think harnesses parallelize work automatically because they’re more likely to hit session limits, and they consume more tokens in message-passing, coordination, and conflict resolution than when serial.

I think this is a plausible explanation, but I still find it unintuitive. Wouldn’t you expect the agent eat up more tokens to do 10 tasks in the same session/context than to have one supervisor delegate to 10 tightly-scoped subagents?

But even token efficiency aside, I imagine there are lots of users like me who don’t run up against quota limits often and would happily trade tokens for faster execution (in wall time).

score 1 · canonical

I think we’re probably more in agreement than you think.

Like the Sandi Metz talk, I don’t think these are rules that everyone has to follow strictly. These are rules of thumb that are meant to help bloggers who have trouble finding people interested in their writing. As bloggers gain more experience, they can develop more nuanced decisions on my guidance.

we have so many spaces where the code we write or emails we send or takes we have must fit within the confines of some extant style guide or database column, but my claim is that your blog is your space to be yourself and there’s nothing forcing you to write a blog “that performs well.”

This is true, and if someone is blogging for the joy of blogging and has no interest in what happens after they hit publish, then this is not a useful post for them.

The vast majority of bloggers I talk to do care about what happens after they hit publish, though. They don’t aspire to be famous, but I rarely see bloggers say that they truly don’t care about whether anyone reads their writing. They’re interested in avoiding mistakes that drive away potential readers.

i find that people who vocally prescribe rules for others, be it in fashion, writing, art, or other avenues of self expression have lost sight of where they have ended up. for the rest of us, it is good to remember that we’re grown ups, we know more, and we have enough experience to gradually develop our own practice in a way that enriches us best.

I’m having trouble understanding what your position is.

If someone is experienced in an activity, they shouldn’t share what they’ve learned about that activity? Or are you specifically against teaching in the form of “rules?” Does Scott Meyers require some self-reflection in your view for writing rules like “Use const wherever possible” in Effective C++?

score 1 · canonical

Thanks for reading!

Do you have any tips on creating/thinking about the list of things your reader might know // how you’d determine who your reader is likely to be?

I personally imagine writing for my past self, before I started learning about a topic. I know other bloggers think of one of their friends and base it on conversations they’ve had with the friend.

I also find it helpful to write out an explicit list of things the target reader would know and what they don’t know. You don’t have to capture every single term you use in your article but setting the baseline like that helps you make decisions about other terms. Like, “I’m assuming the reader has never heard of git, but I tell them to use dig, so I should probably explain dig.”

I think writing out a list is also a good opportunity to check whether you’re writing for a sensible target audience. I’ve worked with a lot of bloggers who are like, “This post is for web developers with 2-3 years of experience,” and so I ask how much would have to change for it to be accessible to beginner developers, and they’re usually like, “Oh, huh. I’d only have to add two sentences to explain two terms.” So, I think deciding which terms you assume the reader knows can be an iterative process as you decide what kind of article you want to write and what kind of reader you hope it benefits.

score 5 · canonical