I read through this to see if my AI cynicism needed any adjustment, and basically it replaced a couple basic greps and maaaaybe 10 minutes of futzing around with markdown. There's a lot of faffing about with JSON, but it ultimately doesn't matter to the end result.
It also fucked up several times and it's entirely possible it missed things.
For this specific thing, it doesn't really matter if it screwed up, since the worst that would happen is an incomplete blog post reporting on drama.
But I can't imagine why you would use this for anything you need to put your name behind.
It looks impressive, sure, but the important kernel here is the grepping and there it's doing some really basic tinkertoy stuff.
I'm willing to be challenged on this, so by all means do, but this seems both worse and slower as an investigation tool.
The hardest problem in computer science in 2025 is showing an AI cynic an example of LLM usage that they find impressive.
How about this one? I had Claude Code run from my phone build a dependency-free JavaScript interpreter in Python, using MicroQuickJS as initial inspiration but later diverging from it on the road to passing its test suite: https://static.simonwillison.net/static/2025/claude-code-mic...
Here's the latest version of that project, which I released as an alpha because I haven't yet built anything real on top of it: https://github.com/simonw/micro-javascript
Again, I built this on my phone, while engaging with all sorts of other pleasant holiday activities.
> For this specific thing, it doesn't really matter if it screwed up
These are specifically use cases where LLMs are a great choice. Where the stakes are low, and getting a hit is a win. For instance if you're brainstorming on some things, it doesn't matter if 99 suggestions are bad if 1 is great.
> the grepping and there it's doing some really basic tinkertoy stuff
The boon is you can offload this task and go do something else. You can start the investigation from your phone while you're out on a walk, and have the results ready when you get home.
I am far from an AI booster but there is a segment of tasks which fit into the above (and some other) criteria for which it can be very useful.
Maybe the grep commands etc look simple/basic when laid bare, but there's likely to be some flailing and thinking time behind each command when doing it manually.
Not the original poster, but that was snark and not meant literally.
Also, building your own plane is absolutely worse, even if you do have expert-level knowledge. That's true for any complex design. Aircraft design, material sourcing, fabrication, assembly and quality control are all very different skill sets, but the real kicker is experience.
The reason why commercial aircraft are so safe is a lot of work goes into investigating and understanding the root causes of accidents, and even more work goes into implementing design fixes and crew training.
The problem is that the system incentivizes incompetence. The mechanics who are paid a skilled wage, take their time, and double check to make sure they are not missing anything show up as big red problems on the beancounters' spreadsheets and get optimized away.
The system can make up for this in other ways like repeatability of processes, redundancy, etc. Which is why commercial aviation is safer than general aviation, and also why I specifically worded my comment as being about the same model of plane - ie if instead of building your own experimental-class kit plane, you hired it out to a liability-limiting company hiring minimum-wage workers to follow the directions. I'm guessing such a thing is illegal per FAA regs, but that kind of proves my point.
For another example, have you experienced the medical system lately? Doctors are generally smart people, but that intelligence is squandered by having their attention smashed into 10 minute chunks, with the entire rest of the system revolving around blame passing - the end result is a lot of smart and well-meaning people ending up grossly incompetent through emergent effects. I would much rather be able to go to a doctor and trust whatever answers they gave me rather than having to do my own independent research and advocacy to drive the process. But that is not how the system we have works.
I don’t even disagree with you about the system incentives. I hate capitalism just as much as you!
But I still trust the institutions around me to keep me safe. Obviously that depends on where you live, I wouldn’t feel the same way if I still lived in Brazil.
Last time I went to a doctor was about 3 years ago. They diagnosed me in 5 minutes, and took another 10 to treat me and write me a prescription. It was great, I loved it.
Sounds like you have this trust issue with lots of different areas of your life, it might be worth reexamining your own perspective. Or maybe you just have to move to somewhere that you do trust.
I'm glad for you that you've had good experiences so far! "Diagnosed me in 5 minutes" doesn't sound like anywhere near a complex medical issue though.
I certainly keep trying to obtain good results from the system, ie extend trust, but situations routinely run aground. Can you really say it's a "trust issue" when the problem is that I dig into details of situations and repeatedly discover how so-called professionals abjectly drop key issues on the floor?
Latest example: I need a new dishwasher. I should be able to read some reviews, spend $1k, and get the problem solved, right? Guess again - first delivery, a dent (crease) in the tub from the thing being slammed so hard that its plastic frame deformed and pushed up into the metal tub. Second delivery - loud noise from wash motor. I try to engage with warranty service figuring I'd be fine with them swapping the whole pump assembly. Nope, the guy that comes can't even be assed to do his job either! "Oh that's normal so there is nothing to fix, this is a good model, you should keep it". Third try, wash motor sounds a little better but still has a problem. The third set of delivery guys didn't even take away unit #2 for the exchange (even though I even pushed back when they said someone else was going to come later). I had wanted to simply pay money to solve the problem, but instead I'm left with two noisy dishwashers and a big ole project in my court. (do I keep pushing this exchange button? do I just order a new pump assembly and fix it myself, considering the bonus dishwasher compensation for that? do I say fuck it to the whole brand and rethink the purchase decision?)
Sure, I could drop my standards here, check out, and stop caring about the details. The dented tub probably wouldn't leak a decade down the line, the loud motor isn't really that big of a deal if I only run it overnight, and if the motor needs replacing in a few years it's only a $200 repair. But should not giving in to this "best effort" service (after paying $1k) really be considered a "me" problem? It seems more like an economy problem, with me only being exceptional for noticing, having some expertise on how these things should function, and having the willingness to push back.
(although I am thankful that the thing in the front of my mind that I'm frustrated with is an appliance rather than dealing with the medical system again)
See my other follow up comment ("same model"). Medical device software development feels much closer to homegrown (or worse) than aeronautical engineering.
Why do you think a random person, who is VERY passionate about something, as to invest all the free hours in life to do something, is less skilled that one who just does it because is needed to survive?
Sorry. I would be much more inclined to have something made by somebody passionate about it, as done by some guy that received hopefully some kind of instruction on how to do things and was then left alone.
In this context (GA) we are not comparing Airbus/Boeing with a garage build. We are comparing some small company making 2 seaters with your hangar and maybe 10 certified aircraft mechanics that will help you a lot on the process.
And why do you think pathos arguments are logical? Granted, they didn't cite them, but assuming it is true, empirical studies showing the accident rates are the logical point from which to draw conclusions. What you would like, how you and others feel about it, and what you would expect are meaningless.
You're also equivocating. They made it extremely clear they are referring to hobbyist and other such groups with vague or unknown qualifications; whereas, you go in and make stipulated claims about small businesses with certified mechanics, etc. These two are clearly not the same category, making your argument non-responsive. It's also contradictory in terms of discussed liabilities and such, as the small company, and its mechanics, that whoever worked with, would have liability as well, as opposed to the "random git repo".
You write that as if you have ample experience with codebases of medical devices and I'm going to take a stab at this and say that you don't. Prove me wrong.
Same. I like the type hints -- they're nice reminders of what things are supposed to be -- but I've essentially ~never run into bugs caused by types, either. I've been coding professionally in Python for 10+ years at this point.
It just doesn't come up in the web and devtools development worlds. Either you're dealing with user input, which is completely untrusted and has to be validated anyways, or you're passing around known validated data.
The closest is maybe ETL pipelines, but type checking can't help there either since your entire goal is to wrestle with horrors.
“The user can choose between starting their new policy on the first day of employment, the first day of the fiscal year, on a specific date, or some number of days after their prior policy expires. If they choose the first day of the fiscal year, the user must specify when their company’s fiscal year starts. If they choose a specific date they must choose a date that is after the first business day of the next month and no later than December 31st of the year that month belongs to. If the user specified some number of months after their current policy expired the user must provide a policy number and the number of days no less than 1 and no more than 365.”
Type validation can help with some of that but at some point it becomes way easier to just use imperative validation for something like this. It turns out that validating things that are easy is easy no matter what you do, and validating complex rules that were written by people who think imperatively is almost impossible to do declaratively in a maintainable way.
Even excusing that daemon, here is a list of processes which have attempted to contact Apple in the past 24 hours, according to Little Snitch. I am certain this is not even a complete list, because macOS is closed source and likely can bypass application firewalls altogether:
Again, I have never used iCloud/Apple services, turned off all available telemetry options and did not open any Apple applications while all this took place (I only use Firefox and iTerm). Almost all of these processes lack a man page, or if they have one, it's one-line nonsense which explains nothing. This is beyond unprofessional.
The scheduling shouldn't be the 5th largest consumer of CPU. The question is what is it scheduling. Collecting data about user behavior would be a background task, you know..
Most of them are, yeah. There are a lot of Idea Guys on here, who are in love with the idea that they no longer need effort or skills to Create Their Vision. If they can just prompt hard enough, success will come rolling in.
Did either of you read the article? You seem to be arguing against a point it doesn't make. Tools like Claude Code are entirely capable today of one-shotting tiny bespoke web apps that do a narrow set of things for an audience of one.
The article isn't talking about "large brownfield projects" or people wanting "success [to] come rolling in". It's about people making little apps for themselves, for personal enjoyment, not profit.
Not true. Border agents were always tinpot dictators, but they're awful these days.
I personally know white collar professionals who were turned back at the border for having a B1 visa instead of a TN visa for temporary work, while another person in their group was let through with no problem. They've been using the same kind of visa for a decade before this. And they were white dudes from Canada.
My Mexican friends said they're not stepping foot in the US for the foreseeable future, and I don't blame them. They have families and can't risk being disappeared and potentially killed because of some idiot border agent.
The Python leadership group is friendly towards beginners and people with non-traditional backgrounds, and they actively reach out and recruit those folks to participate. It's also one of the least toxic developer communities I know of.
That matters a lot in a field like data science or cybersecurity, which has a lot of people who don't have traditional SWE backgrounds.
The language is pleasant and readable to use and has a lot of features built in, which Perl did not, but I think that's much less critical than the network effect of being a (1) welcoming and pleasant community who (2) actively recruits new fields of people.
reply