danielbarla

I see these more fundamentally as arising from whether a good feedback loop is possible or not. The same is fundamentally true for human experts; you need validation that your work was good or not, or you cannot improve.

With code, it's fairly easy to have unit or end to end tests. That can be relatively objective and deterministic, and pretty quick.

Architecture is one layer removed, and far harder to test. You're stuck with things that are at least an order of magnitude to prove autonomously; will it scale in terms of performance, will the abstractions be a good fit for future requirements, etc. Testable, but much harder.

Design; well, there's possibly some metrics by which you could measure UX (attention for the viewer, contrast ratios, number of actions needed to achieve something, how far the mouse had to travel), but really the majority of it is subjective. "How does the human feel when viewing this" - I suspect almost by definition needs either a human, or a very good model of it. I don't think we have the latter, making this pretty hard.

> You can, but nobody does.

I for one, have far more rigorous quality checks in my hobby projects (where AI coded), than I ever could justify when I hand-coded them.

I'm not claiming to be everybody, but surely a good portion of the population are using these technologies similarly.

Phrasing it as a UI shortcoming is underselling it. The search mechanism itself was dumb[1] and therefore the user had to be more intelligent. As an example, knowing synonyms for what you were searching for, in order to disambiguate the query from a more popular term that would likely dominate the results but was irrelevant for you, was a major skill.

[1] Yes, we can argue that LLMs are dumb, also. But they magically do "get" your intent with your prompt, unlike the search engines of yesteryear.

Indeed. I'm honestly shocked that we're still having to evict bad query plans in 2026. And "you changed a variable", ha, that sounds like an actual reason. How about "data statistics were automatically refreshed and you hit some magical undocumented heuristic threshold, an the query that ran in 35ms yesterday now takes 45 minutes. And we can actually tell you this because we have the data, but decided to let you find out manually, instead."

I've literally been saying "I can't believe the date is X and we still have to put up with this" for around 25 years now.

>they do track object across video frames which for me it's the minimum a system like this should do. All previous FSD systems are just Fake self-driving.

I'm pretty sure Waymo does this. It's hard to imagine not using this information, it's basically 50% of what I do when driving (e.g. that kid that was running on the sidewalk suddenly disappeared from view in an unexpected way; I wonder if he's about to cut across the street, etc).

> is that not enough? what else do people want?

I've bounced off NMS several times, despite having tried quite hard to get into it. So, my take is based on ~20 hours of early game, from 2025ish. Probably the best parallel I can draw is to Elite: Dangerous, which I played for several hundred ours before getting bored of it. What I find lacking in both games is that there's no _meaningful_ emergent gameplay that spawns off the procgen. Everything is wildy different on each planet, yet very same-y, because none of the differences really play into each other. In both, one can discover random biomes and creatures, but it's not like they are inherently interesting, they are mostly just following the same scripts.

Contrast this with something like Dwarf Fortress, which takes a slightly more in-depth approach to procgen, one where things have history, and in many ways, things "make sense". Or Rimworld, where the game tries its hardest to conjure up a narrative for you, dynamically, but still in a sandbox. So, basically, NMS is a perfectly functional sandbox game, and I have huge respect for them for keeping on working at it for so many years. However, it doesn't scratch my "hardcore simulation / mastery" itch, and so far I haven't been grabbed by any of its story generation. It just felt like a long series of fetch quests and "collect them all" games.

In general, I completely agree with you. Personally, as a human: I have fallible memory, beliefs that are not as sound as I may believe, and all of those notwithstanding, I may not correctly reason, even if my "inputs" were perfect. Let's not even get into cognitive biases, etc. I am acutely aware of my, and other humans' fallibilities.

I think when people are irritated by the "hallucination" aspect of LLMs, they are often running into something of a slightly different nature. I mean, firstly, there seems to be a higher-than-normal ratio of these "little mistakes". And secondly, there are some pretty odd ones - e.g. in my team, Claude regularly just straight up makes up Jira ticket numbers, and then refers to them with high confidence. I guess what I'm saying is that a human would probably not just make up an id, and run with it (they may be off by one, or mix two up, etc). In my opinion, these can be successfully treated, but I guess you can never fully eliminate the tendency that irks people.

Aside from the other reply, pi makes it _really easy_ to build your own usage tracking and limit machinery.

I would normally advise against such efforts for a variety of reasons (such as inaccurate tracking, etc), but specifically under pi, this mechanism has been extremely well behaved and accurate for me.

Several people have given the opinion in this thread that DF is basically immune to having a tutorial; I am trying to figure out why. I don't see a tutorial as an end-to-end blueprint of everything the game is capable of. It's just a series of info, hints, and follow-along instructions, possibly with railroaded / simplified UI. I mean, the first time I booted up DF with no context circa 2013, it took me a few minutes to figure out that we're seeing a top-down view. I certainly would have appreciated a hint or a tutorial (though I do not necessarily fault the game for not having one; it's _not_ a game that is aimed at onboarding heaps of fresh newbies by the truckload each week).

Yes, I think we're in agreement overall.

I think the hate typically comes from the mis-application of various tools; overly complex solutions for simple problems. There's nothing wrong with patterns per say, or frameworks that have complex building blocks for every conceivable permutation of relatively common issues. However, when people reach for unnecessarily complex tools out of laziness/habit, some hate is justified. I suspect this is very similar to what you're saying, and of course, it's not OOP specific. Functional languages have their patterns too, but instead of a famous book enumerating them, we have a famous blog post stating that functional languages are so awesome that they don't need patterns (or at least, this is how most people seem to interpret it).

> But I also spent some time with modern Spring Boot, and I get where the hate is coming from. There is a lot of complexity that just doesn't seem necessary at all.

I'm sure there is some (very high) level of inherent project complexity where the massive number of abstractions and tools that Spring Boot gives you (and forces on you) actually starts turning into a net positive... And I'm also fairly sure that most projects don't quite break even, and would be better off with something more lightweight.

I agree with Gabe Newell's take, however it absolutely is _also_ a money issue (which Steam went a long way to alleviating with regional pricing). Median PPP ratios for other continents are: Asia ~2.76x, Africa ~2.76x, Europe ~1.58x, etc.

Currently new AAA titles debut for 60 to 80 USD. I'm wondering what the piracy situation would look like in the USA, if this was, say 160 to 220 USD; because that is what paying US prices feels like to the median person.

Aside from the skills mentioned, I recently found that asking the model to create a simple HTML presentation to walk you through it's proposed design can do wonders. Fine tune your prompt to your liking (language style, what to include, what not, etc). Then, read that thing thoroughly, and keep asking questions and iterating on the plan until you fully understand what's about to be built, and are happy with it.

> Tbh, I think it's a terrible question.

I think it's a terrible _interview_ question, partly due to it's "IQ test" nature, but mostly because it's a pretty hard problem and I really wouldn't expect someone to be able to just spew out a solution the first time they hear it.

Of course, the process would be more about the interviewer observing your thought process, seeing how you develop a notation for solving this thing which you almost certainly don't have any pre-existing notation for, etc. I just really wouldn't expect much progress in the space of 5 to 10 minutes, although I am basing this on the "13 marbles" version of the problem.

I've been using a plain ASD-STE100 prompt for a little bit now, and while I somewhat prefer the text that agents produce while being told to use simplified English, it does not necessarily change the overall structure of what they write. It just now tends to have shorter sentences, and fewer drumroll, pitch-deck / vacuous section titles like "the finding that pivots the preposition", etc. Overall a significant quality increase in my opinion, but nothing earth shattering, and doesn't fully solve the problem.

Indeed, I suspect the approaches/algorithms for solving a Rubik's cube "compress" a lot better than trying to distill the entire search space in order to be able to predict the exact next move.

I see this trope fairly often, i.e. the assumption that an LLM would need to have been trained on <exact thing it is being asked to solve>. Now, while I do have a moderate amount of background in AI, I am definitely not an expert on LLMs as such. I would be interested to hear someone's take, who does work actively in LLM research. Can they generalise "well enough"? They certainly seem to be able to do so, from my anecdata, and I don't believe "training explicitly for every possible scenario" would have scaled even to today's state.

> Somebody on the forums posted a tool that converts generic joystick axes to keypresses, but not sure how well that works.

I don't have a TSC-X, but did frown a little bit that there wasn't a generic support for controllers with plain axes. I have a VKB STECS, and basically a 85% of train sims need a workaround.

The good news: Claude Code threw together a working prototype for me in a few minutes that had me mapping the two axes (power / brakes) perfectly. It's really a low barrier to entry these days. Can confirm that this game is amazing with it.

Somewhere around the initial ChatGPT release, I was experimenting with having the model draw various types of ASCII art. I believe the first time I ran into a doom loop might have been when I asked it to draw a giraffe; it started with a head, then a line of | | for the neck, and then it just repeated that, ad infinitum. I am not really sure what the cause was, but clearly it wanted to make a long-ish neck, and ran out of a way to determine how many lines it had already produced, so just kept going.

If we're going with a math analogy, I guess it's a bit like teaching them integrals in 3rd grade. You can do it, they probably have the raw IQ for it. But they won't really understand and appreciate it at a deep level (this is even a problem for people when they encounter integrals at the end of high school / early uni).

Novels like these need some life experience to really shine. A 13 year old isn't going to go "how does this writer see so clearly through so many of life's finer details", because they have never experienced 90% of what's being talked about.

Aside from the pragmatic reasons, I think it's a good idea to separate out cardio, muscle-building, and flexibility into its own separate categories, and ensure you consistently dabble in each. Obviously there can be are overlaps, but this taxonomy ensures a good balance.

I don't see this as a binary thing. Legally we tend to draw a clear line between child and adult for pragmatic purposes, but I don't think my responsibility of intent disappears just because someone hits a magical number. I have steered clear of various gambling / "gaming" jobs which have had silly high salaries as a result; I don't in any way want to participate in things which are meant to play the weak points of the human psyche like a harp, for profit.

To be fair, this topic seems to be quite divisive, and seems like something that definitely should be discussed during an interview. Who is right and wrong is one thing, but you likely don't want to be working for a company who has an incompatible take on this topic to you.

I find the chess comment fairly weak; while chess is a "perfect information" game (in that you have full information about every piece on the board), at every level that chess is played at there will be unknowns about the player you are facing.

At lower levels, you can certainly play sub min-maxed moves that are more likely to confuse or pressure a weaker opponent. You could call these bluffs. Opening theory is also a wonderful game, you're essentially playing a game of "let's bet that you haven't prepared this particular set of lines as well as I have" with the opponent. The same thing goes for game styles, open vs closed, etc.

Last but not least, playing a perfect game of chess is so far out of the realm of possibility for humans, that the entire "there's always a correct move" is completely irrelevant. We are now at a point where 7-piece endgames have been completely solved, and it involves 4.2×10^14 positions. Good luck memorizing that, and the scaling from there on out is not pretty.

In this sense, chess occupies a very interesting spot; somewhat calculatable for a human (and yes, tactics dominate up to say 2000 ELO), but there's plenty of room for creativity and strategy also. It's also played at an insanely high level, which makes it a worthwhile challenge and time investment. What it does NOT have is randomness. I often wonder what the competitive landscape of a chess variant involving some randomness would look like, or if it would fundamentally change the nature of the game.

I completely agree, it's fantastic both for general fitness, and as a way of exploring. I do this on a smaller scale on a daily basis, walking the 8-10 km route to or from my work (when we have office days). This is walkable in about 1.5 hours, public transport would get me home in around 45 minutes, so I am not really investing much extra time. Varying the route slightly keeps things interesting, and you get a surprising variety with small (1 to 2 street) changes.

Another favourite of mine is cycling around a neighbourhood to get to know it; you get a totally different feel for things than from a car - things typically go by slower, and you are somehow just far more able to observe things.

Availability is definitely a factor, but I feel that a far more important aspect is that a YouTube feel is personalised. It's A/B testing you for weeks on end, and has a pretty good idea of how to get maximum engagement. TV was never this targeted, nor was there feedback to ratchet up what it suggested to you.

Because obviously, we can be trusted completely!

> Why not!

Responsive layout would be the biggest reason (mobile for one, but also a wider range of PC monitor aspect ratios these days than the 4:3 that was standard back then), probably followed by conflating the exact layout details with the content, and a separation of concerns / ease of being able to move things around.

I mean, it's a perfectly viable thing if these are not requirements and preferences that you and your system have. But it's pretty rare these days that an app or site can say "yeah, none of those matter to me the least bit".

Godot is a great engine, and .Net support is very good. You can't go far wrong with it, especially for small 2D games.

I think 5-10% is pretty good, it probably means that the codebase is mostly understandable and maintainable. I have definitely worked on some which were full of little traps and landmines just waiting for eager do-gooders to step on, which was sadly a self-fulfilling prophecy for the app.

Very interesting post, thank you!

I'd also be curious to know the following: how many new errors or regressions were caused by the bug fixes?