hathawsh

We might be at the beginning of a golden age for inventors. Robots of all shapes and sizes, with highly variable price points and specialization. Still, the first challenge for R2D2 is probably stairs.

I've found that AI-written text consistently fails to comprehend the context. It either assumes there's little context available, so it writes too much, or it assumes the reader has all the context, so it writes too little. Human-written text has those problems far less often.

I hope the LLM builders are paying attention and they add some kind of step in the process that causes the system to reflect on its assumptions about context and work to refine its understanding of the context before generating output.

Or it could indicate that more resources have been put into cryptography than most other math fields, so AI hasn't yet caught up to our progress.

Well, to a CMS developer, a lot of problems look like CMS problems. I know because I used to be paid to think like that. Portfolio website? CMS. Support ticket database? CMS. Newspaper layout system, intranet portal, blog, restaurant menu, identity verification system, etc? Yeah, we built all of it using a CMS. It made a lot of sense at the time because HTML and CSS were young and difficult to wrangle. Today, most of the reasons to build or use a CMS have evaporated.

That's a great perspective. Come to think of it, I wonder about the quality of the build123d code it generated when I vibe-coded a prototype complex assembly. Now I'm curious.

Yes, but that's in psychology. Different fields have their own terms without causing too many issues. Communication breaks down when we reuse terms for different things within the same field.

The LLMs have been trained on all kinds of code. They often choose a hacky solution because that's what they encountered in similar situations in their training. It generally only takes a few words to guide them to a better design than what they saw in the wild, but if you don't guide them, they follow the herd.

I will be unhappy if they steal the meaning of NLP. NLP means Natural Language Processing and LLMs are only a subset of NLP. I would have concerns about anyone who claims to have technical knowledge and asserts NLP means "natural language programming." Those are the same kind of people who might suggest Lisp should add mutable objects.

Growing up in my family, "hot sauce" was Pace picante sauce / salsa. I think that's true for a lot of America also.

https://theempirebuilderspodcast.com/244-pace-salsa/

Tabasco is OK, but Pace is the sauce that we always consume before it expires.

There is a large market for fine art. For everything else, there are printers and copiers. It's very difficult to maintain a career of painting and selling works for $100k. It's much easier to maintain a different career that happens to include printing things with little effort using an ordinary office printer.

That's where software development is going: of course you can still code by hand, but it will be increasingly difficult to get paid to do it. People will have a somewhat different career that happens to include coding things with little effort using an ordinary coding agent. Companies won't be able to afford to pay for expensive coders anymore (because otherwise their competition, who is getting along fine with a small software development budget, will take all their customers.)

I thought the same until I had kids. I realized that kids generally want to learn, but they usually think what the teacher is telling them is not the best thing to be learning. No technology, not even AI, overcomes the student's ego; in fact, technology rarely even tries to address that issue. I think this often applies to adults too. Teachers continue to be necessary to help students achieve an appropriate level of humility to be able to learn.

You clearly have enough humility in your field to be able to soak up knowledge. That's great for you! I suggest your dedication is rare and teachers continue to be essential for society.

While I agree that a coding model (such as Opus) by itself tends to act very shallowly, when it's driven by a harness like Claude Code, the combination seems to be a far more general thing than a LLM. It's capable of consistently making excellent data structure and architectural choices over large code bases. It imitates thinking about anything and it can drive itself for hours.

Honestly, if I simply fed it a sense of presence (I would repeatedly tell it what's going on right now and ask it to react if it thinks it should), it would feel eerily like AGI.

I try to remember to say something like this: "This is an interesting solution, but I'm not confident we have a shared understanding of the problem we're solving. I have a few guesses, but I don't want to bias the discussion, so please restate the problem without suggesting a solution. If it's too difficult to explain, I can suggest what problems I think you might be running into." I wish I always remembered to say that. :-)

Thanks for the correction. In my own words, to commit to memory: once "the cat is out of the bag", the point is not that it's difficult to put the cat back in, the point is that we now know it's a cat and putting it back is not relevant.

"The cat is out of the bag" doesn't mean the cat is gone, it means the cat has freedom. I read the GP as meaning "no single company can currently control the trajectory of AI." That's not unfortunate -- that's great!

Last time I joined a discussion about US health care on HN (years ago), the picture looked bleak: secret chargemasters in every hospital, no realistic way for patients to compare options because the options are too complicated, and lots of people still falling through the cracks.

I can't help but think the picture looks very different now. There is growing awareness that chargemasters can not be secret. LLMs offer a way to overcome the complexity of choosing options, opening doors for real competition. I sincerely hope we can take advantage of this changed landscape to provide better care. Universal health care remains an interesting possible direction, but I now have hope that we can do a lot better within the existing laws.

That's fantastic. I have long, winding trails near me also and one of these days I also want to start prompting a coding agent on my bike with a headset. Do you recommend any particular type of headset?

Edit: never mind, I see you already suggested the Shokz OpenComm2 in another comment. Thanks!

Yes, but null pointers are so pervasive in C code that we really can't afford to put assertions everywhere. It's often better to let the app crash on violations.

Iroh looks very interesting!

How current is the PyPI package? https://pypi.org/project/iroh/

If I were doing a code review, I would probably accept the code either with or without the assertion. The context of curl_getenv() makes it clear that null is not acceptable. If the author of curl_getenv() had evidence that callers are frequently breaking the contract by passing null, then perhaps the assertion would help shed some light on violators. Otherwise, I would expect everyone to play by the rules, making the assertion unnecessary.

When I tell my coworkers to stop using AI to dress up their words, it's not because I care about human effort. The problem is that my coworkers often start with incorrect assumptions, and AI is good at amplifying bad assumptions and making them sound plausible. I have to spend extra time guessing at what the author originally wrote and then address the partly-hidden original points rather than what the AI generated. Give me your spelling errors, your grammar, your mumbles, your incoherent streams of thought, your doubt and uncertainty. Those things are extremely important, yet your robot obscures them.

Strangely, I've also observed that some customers respond very well to words dressed up by AI, even if the words oversimplify the truth. Now I'm working to understand why they want that. Are my customers not swimming in AI slop like the rest of us?

BTW, this doesn't mean I'm anti-AI. AI coding is an incredible superpower and I use it constantly, but it seems to me that AI coding works because code expresses the minutiae that is rightfully omitted from most other communication.

Sorry to break it to you, but on that timeline, the good things got poisoned. IBM enhanced Lisp with Enterprise Ready features like Spreadsheet Macro Builder, Microsoft took over development of Smalltalk and morphed it into BASIC 2.0, and the HURD community lost a bizarre copyright lawsuit. Fortunately for those folks, an intrepid hacker in the 90s saw some of the interesting ideas in MS-DOS and rebuilt it as LS-DOS. Today, most of their servers and mobile phones run LS-DOS or similar.

Your HN account is too new for me to be sure whether you're being sarcastic or not. Perhaps you know, or perhaps you don't, that all code is machine-translated, even assembly language. None of it is perfect, but it's not garbage. Today's AI merely provides a new level. It's a weird, non-deterministic level, but hiring an employee to write code for you is similarly non-deterministic.

These are also the markers of human journalists who write daily. Journalism is the reason AI acquired these habits. Gemini says this article is probably not generated by AI, particularly because it has original quotes.

https://gemini.google.com/share/ba48849a15a9

I actually agree with this. Product designers and product managers are often essential and sometimes they do up to 99% of the work of figuring out how something should work. To accomplish that, they often do things well outside the role of a software developer. On the other hand, in my experience, only someone with a software development mindset seems to be able to complete the last 1% (or 10%, or whatever) that reveals and resolves certain kinds of logic issues.

Isn't that interesting? The job of exploring a theory or model to such an extent that it can be expressed in computer code always seems to fall on the shoulders of a software developer. Other people can write specifications and requirements all day long, but until a software developer has tackled the problem, the theory probably hasn't been explored well enough yet to express clearly in computer code. It feels like software developers are scientists who study their customers' knowledge domains.

Is that actually true, though? Even though it's not really my job, I find myself debugging certificates and keys at least once a month, and that's after automating as much as possible with certbot and cloud certificates. PKI always seems to demand attention.

This is an amazing resource. It was difficult to appreciate what this resource was for until I tried to create my own boards based on an ESP32. It's not really difficult to build around ESP32, it's just that I don't know what I don't know. With starting points like these, I can start with a lot more confidence. Thank you!

You hinted at an aspect I probably haven't considered enough: The code I'm working on already has many well-established, clean patterns and nearly all of Claude's work builds on those patterns. I would probably have a very different experience otherwise.

I love the IDE skins analogy. Very true.