arscan

Hey, I worked there in high school as an intern! But just on the intranet, which for its time (‘96?) was really quite good.

I would think this type of behavior (ugh, or dare I say default mindset) by consumer-facing LLMs will always be desirable for ‘hard’ problems (things previously unsolved) because it’s a bit like having saftey mechanisms in place to prevent hallucinations for users incapable of verifying correctness of the output. You’ve got to do a little work to prove you understand that it’s hard but it’s still something the LLM might be able to accomplish.

I think fundamentally there is a difference in the definition of “America” at play here too.

> This isn't just a new feature — it’s a new way to build software

I imagine they smushed 2 of the most well-known AI-isms together on purpose here to troll the AI-weary. Might as well I guess, if this type of language annoys you then the whole feature probably will too.

I was very into PC gaming during this era, and ran a decently sized gaming news website, and Carmacks use of .plan as a one-to-many microblog was really a decade or two ahead of what became mainstream twitter’s primary use case. None of the games I ‘covered’ had enough .plan activity by the developers to warrant an automated script, but I remember being really impressed by the sites that did it (usually id-centric ones like blues news or stomped I think?)

Such good memories. I think my whole career has been built around chasing the feeling I had when making that gaming site.

Well, yeah, it does.

Effective and clear communication is really important and often really hard for engineers. It is said that one goal of stackoverflow was to help programmers learn how to write through practice, as it’s both very hard and very critical to their effectiveness:

https://blog.codinghorror.com/how-to-write-without-writing/

But programmers are also lazy*, so why put in the effort to improve when you can just get llms to do the work for you?

* not necessarily in a bad way

Not a substantive comment on content but hopefully constructive feedback on presentation:

Holy moly, on mobile I was trying to read the example console screenshots/snippets and then it would just unexpectedly change. Took me a little while to figure out it’s some kind of carousel for the examples, and not more screenshots/snippets loading and pushing down content (or me going crazy). Please don’t do this on mobile sites, just let me scroll through the examples!

There’s the chance that you notice patterns emerging that become quite grating after listening to it for awhile, in the same way that LLMs tend to have subtle but eventually noticeable patterns in writing. Some people care, some don’t.

Will that be the case for the music that you are listening to that’s generated by AI, maybe? Maybe not?

By highlighting that the music predates generative AI, at least that’s something that you won’t have to think about.

At least, that’s how I read it. It’s like a ‘no artificial sweeteners’ label on a drink.

Who is going to end up capturing all this value being generated is going to be very interesting. Back in 1980, who’d have thought MS would capture the majority of the value from PCs over the next 3 decades, and not IBM?

Linus Tech Talk (LTT) did a whole series on doing this on the pool at the channel hosts’ house. Extravagant home upgrades are a frequent topic on that YouTube channel… business expense write off yada yada. My general takeaway was, yikes, all that piping and infrastructure would be a nightmare to maintain and will likely just be closed off whenever an issue comes up (or he sells). I’m no expert, but I am a home owner, and have come to form a deep appreciation for maintaining simplicity when it comes to the operation of your house.

Perhaps this is a controversial take, but I believe security runs in opposition to many other quality attributes (i.e., usability, interoperability). Not to say that there aren’t designs that could exist that are optimal for security and these other quality attributes simultaneously, but security massively reduces the potential design space to the point where there is no rule that at least one of these optimal designs must exist.

Quality is an opinionated measure of a design, which involves tradeoffs based on your goals, values, etc. Valuing security very high, even at the expense of every other aspect of quality if necessary, is completely understandable. But that’s a value that isn’t universally held.

If your point is that security should at least be considered when measuring quality these days (like, it’s a top 10 ‘ility’ at least): fair enough.

Quality is an opinion based on your perspective. The author articulates quality from their perspective, which has value, but isn’t universal.

I remember Matt’s Scripts Archive as an absolute gold mine for learning how to make web applications through example in the pre-PHP days, which was pretty challenging when all you had to work with is CGI and maybe SSI if your hosting provider was particularly advanced. It’s what got me started as a web application developer 30+ years ago. I guess I probably learned about security the hard way by following his examples. But it got me headed broadly in the right direction I think.

I remember being very proud of how I extended his forum software to support threaded messaging and pagination.

Applications like Photoshop will one day be regarded as the castles of this era. Technically impressive, but economically/politically irrelevant. While they could be reproduced at a fraction of the cost there just isn’t any point to it because there are much better ways to allocate capital.

I had a similar experience. Back then gamers were an outsized part of the online community, so if you wanted to build a site that got some actual use and engagement, building to that audience was a good strategy. And of course, it helps when you are part of that audience.

I learned about this technique from Owen Wilson’s character in the otherwise exceptionally forgettable movie “The Haunting (1999).” Paradoxically, you are the one doing them a favor by effectively giving them permission to ask for help in the future.

I interned at a company called Stratus which did hardware fault tolerant computers in the 80s/90s. I think they called it a “Pair and spare” approach, where every component had 3 copies running and comparing state every cycle. If one component’s state stopped matching the other 2, the failing component would be taken offline and the system would call home for a replacement to be fedexed overnight. I think just about every component was hotswappable too. Pretty cool, but expensive, and other architectures for improving availability, or mitigating impact from loss of availability, won out (except for a handful of exotic use cases).

It does, but according to the article it doesn’t apply to tickets issued by camera.

To perhaps give a little insight into why this is on the front page by someone who upvoted it: I didn't realize it was so open and easy. Now I do. The Golang code simply serves as proof in how open and easy it is.

> Even at the time, the task didn't seem like "enough" for a show-the-world blog post.

Its an old (de facto industry) standard, but maybe more relevant than ever. I'm interested in moving more of my compute usage off-cloud these days, which is why this is of interest to me right now. I suspect many others feel the same way.

Might be a good time to post other tidbits of knowledge you have like this, targeted at software engineers that are starting to get more into infrastructure management. Standards that are ubiquitous and just work are awesome.

It is completely possible that the path that got them to this point was the optimal path given their goals and knowledge at the time. And wildly enough, maybe it was even the optimal path with perfect knowledge of the future as well.

I don’t think people are spending their time on more pressing issues. I think they are just are hooked on an endless stream of content that is built for addiction and is always within arms reach.

I think the idea is that if they cannot perform any cognitive task that is trivial for humans then we can state they haven’t reached ‘AGI’.

It used to be easy to build these tests. I suspect it’s getting harder and harder.

But if we run out of ideas for tests that are easy for humans but impossible for models, it doesn’t mean none exist. Perhaps that’s when we turn to models to design candidate tests, and have humans be the subjects to try them out ad nauseam until no more are ever uncovered? That sounds like a lovely future…

I do love the concept, but a little part of me died each time I came across an article with a very strong AI voice. That just feels antithetical to the ‘small web’ ethos because it obscures the ‘neighbor’ behind it.

Sure but industry cares about value (= benefit - price), not just price. Price could be astronomical, but that doesn’t matter if benefit is larger.

I’ve been thinking about it like this for some time: If the computer is a bicycle of the mind, then the LLM is its credit card.

I’m not so sure an increasingly large context window will be seen as a critical enabler (as it was viewed 6 months ago), after watching how amazingly effective subagents and tool calls are at tackling parts of the problem and surfacing the just the relevant bits for the task at hand. And if increasing the context window isn’t the current bottleneck, effort will be put elsewhere.

Reminds me of the “Google AI Challenge” in 2011 called Ants [1], except the ‘AI’ is implemented using ‘AI’ now instead of human programmers.

I was proud for getting the highest-ranked JavaScript-based implementation, but got absolutely crushed by the eventual winner.

1. https://github.com/aichallenge/aichallenge

> There's a middle ground of "written by human and LLM together".

Absolutely, but I’d categorize that ‘bit’ as the innovation from the human. I guess it’s usually just ongoing validation that the software is headed down a path of usefulness which is hard to specify up-front and by definition something only the user (or a very good proxy) can do (and even they are usually bad at it).

> But I’ll gladly use a tool someone had an AI write, as long as it works (which these things increasingly do).

It works, sure, but is it worth your time to use? I think a common blind spot for software engineers is understanding how hard it is to get people to use software they aren’t effectively forced to use (through work or in order to gain access to something or ‘network effects’ or whatever).

Most people’s time and attention is precious, their habits are ingrained, and they are fundamentally pretty lazy.

And people that don’t fall into the ‘most people’ I just described, probably won’t want to use software you had an LLM write up when they could have just done it themselves to meet their exact need. UNLESS it’s something very novel that came from a bit of innovation that LLMs are incapable of. But that bit isn’t what we are talking about here, I don’t think.

I got the sense that Kidders side was approximately equal to my father’s side, as my father said he provided a lot of information to the author through interviews and was happy with the account that ended up in the book. I’ll see what I can find though.

I ended up working for the lead of the competing team within DG (whose product lost to the book’s protagonist) for many years right after college at a different company he founded. I suspect he has a slightly different perspective on the whole thing, but I never asked.

Sadly my father and many of his contemporaries are no longer with us. But I’m really happy that this book exists as a durable & accurate snapshot of the period. The computer history museum also has a wonderful collection of interviews worth checking out, which includes several of the staff from DG [1]

1. https://computerhistory.org/oral-histories/