qdot.me

@EskoLuontola @GergelyOrosz Yes! New sources of bugs abound! It is a great step in the right direction… and then the physics and politics will hit us once again. Time genuinely is hard. What would be the future offset between UTC and UT1. What do you do with leap seconds and what is the truth. Telco, PV?

@qdotme @GergelyOrosz RFC 9557 was made for this

Actually - no. This is why every and any sane software engineer uses 1) Utc timestamps (no daylight savings) or Epoch time (timezone independent) 2) Date libraries to render dates that take care of local conversion and pesky things like daylight savings

@_virgil19 @paulg There are probably a small handful of libraries out there (sqlite is a good one) that *actually* test the documented scope. Across implicit assumptions, architectures, malformed inputs, etc. Most of the library code out there is just „well, works for me/us”.

@qdotme @paulg fair, deviating from the documented pattern puts you right back to being the first test pilot. the trust was never in the library, it was in doing what already worked

Before vibe coding became a thing, programming was already evolving in that direction. It already increasingly consisted of installing and configuring stuff other people wrote, without reading the source.

@_virgil19 @paulg Deviating from the demonstrated pattern. The documentation is often extremely optimistic and full of assumptions that you don’t know if you’re in violation of. I’ve had HTTP client libraries dislike some servers but not others, compute libraries make memory alignment assumptions.

@sam_kritch Hold a stop sign at a busy intersection :)

anyone know a job where: - I can use my hands a fair amount, rather than my mind (my mind is tired) - I don't have to have any particular skill with my hands (I don't) - I can interact with intelligent people sometimes - It's not bartending, nor art, it is useful

@GergelyOrosz I also have millions of calls at noisefront.com from 2-3 years ago. My customers expect me to give them a heatmap of when they occurred in local timezone for the *customer* or *business* - if most people call at 5pm, you stay open longer.

@GergelyOrosz Actually, no. The real problem is that users expect future timestamps in local timezone - I have my dentists appointment at 10am on October 25. He will expect me at whatever 10am is, locally, that day; not on what my calendar app thought it would be in UTC based on today’s model

@peterrhague Roll the ball 8-10 times. 8-10x the circumference is my engineering approximation.

To go with my 10cm globe I got a 2.5cm marble to represent the Moon (it’s a mm or two too small but close enough) Without working it out, have a guess how far apart I would have to place them to be scale correct?

I’d say anytime you need to navigate the regulators, you should do it the way @boomsupersonic did. Engage in public, with the voters, with the executive, with the legislature. The regulators serve the people, are paid by the people and are checked and balanced by the people. Old nuclear (frankly, old anything that’s publicly traded, next quarter, no real skin in the game by the executives) missed that ball greatly.

Every day I see consultants on here arguing that new nuclear companies need the consultant class’s help to navigate the NRC. No, they don’t. The whole point of NRC reform is to make it possible to deploy new reactors. Just engage with the NRC; don’t hire any of these obsolete grifters.

@octal Conjugation of the application of force. We’ll sh**t. State’ll sh**t. Nobody’ll sh**t.

Crazy going from home (some property crime, murders of gangs/DV, rifle <3m away most of the time, pistol w/ 2 mags when I go out), to China/Japan ("leave phone on table to show it's occupied"), to London/Europe ("no weapon, high property crime, stabbings.")

Exactly. I also formed my pro-life position while being religiously agnostic, as follows: 1) Life of mother takes absolute priority - as does the life of „more developed Siamese twin” etc. No debate. Elective abortionists muddied the water and have contributed to the deaths of mothers who had legitimate medical emergency and were wrongly deemed as seeking elective abortion where illegal. 2) Quality life of a child with genetic diseases or developmental abnormalities. Genuine ethical debate, but I can see supporting it in communities that already support elective euthanasia and under similar medical guidance; ditto post-natal DNR debate - also a solid nut to crack, but again, not the 95% of elective abortions we are seeing in the US. 3) The rest: It is medically interfering with a healthy and natural process. If I show up to a surgeon to have both my knees amputated and replaced with spring-loaded prosthetics because I want to run faster (true fact, Paralympics records etc), I’d be referred to a psychiatrist. 4) I wouldn’t judge mothers who abandon their child after delivery for economic or lifestyle reasons. That’s what we pay our taxes for - to give the kids the chance in foster care etc. 5) I personally would support any mother in my community and would encourage everyone to do the same, if they decide to raise the child. This is also my ethical duty, as is leaving the seat in a public transport to an elderly etc.

You're very wrong. You can disagree with the pro-life position, but it's simply false to claim it's purely religious. My view has nothing to do with religion. It's based on biology and ethics. Biologically, fertilization creates a new, living human organism with its own unique DNA. That's a question of developmental biology, not theology. The real disagreement isn't about whether the embryo is biologically human. It's about if ALL human life deserves legal and moral protection, or just arbatrary classes of it. That's a philosophical and ethical question, not a religious one. My position is that human beings shouldn't lose the right not to be intentionally killed because of characteristics like age, size, level of development, dependency, or location. I don't see a principled way to say those traits determine whether someone's life has value. You can disagree with that conclusion, but calling it "just religion" ignores that many people arrive at it through science, ethics, and secular moral reasoning. I'll even go further and say that as a Christian, the spiritual reasoning against abortion is actually WEAKER than the secular and scientific reasoning.

@number_pizza111 No kidding. With all the bugs and unfinished threads Asahi has, my 5y M1 still kicks butt. And works the way I want it on Asahi, not the way walled gardeners dog wants me to.

Experience with a MacBook after 2 days: MacOS doesn’t work and hates you and actively plots to ruin your life The hardware is so much better than any of their competitors it’s unreal

@_virgil19 @paulg But in either case you might be the first carrier landing, and the thousands of people who succeeded on airstrips are giving you the false sense of security. Most of my library issues were with using it „as documented but differently” - doing exactly the same thing is waste.

@paulg a library survived because thousands of people already hit its bugs before you did. code an agent just wrote for you hasnt survived anything yet, youre the first test pilot

@gnegavin @lauriewired And.. do the LLMs actually become capable of de-novo reasoning (the real joy of programming for me), or merely stay in the drudgery of „and now fill in the rest of the owl”. I know of phenomenal math and biotech discoveries by LLMs, but they tend to cluster around „fill in”.

@lauriewired Maybe a third option: they like computers, just not the layer you do. Nobody mourns hand-writing assembly — that got abstracted and we called it progress. The question is whether reasoning can be abstracted without losing the thing, or whether it's different in kind.

I’m convinced that a large % of programmers don’t actually like computers. As a side effect, are also perfectly happy to throw away their reasoning to a model as soon as they can. I don’t get it, at ALL. Don’t you *LIKE* understanding the magic of the machine? You do realize hand-programming (I hate that I even have to specify hand now) is fun…right?

@SlightlyRelevnt @AndyBeshearKY Just stop electing old people, it’s not like they magically got that way between Election Day and now. Now, we should on the other hand require physical (hybrid) participation in the official buildings at least once a week, with a press conference once a month or lose the seat.

The fact we're even reduced to publicly speculating about our lawmakers' health is a damning indictment of American politics. How exactly is an 80+ year-old supposed to grasp the desires, struggles, and pace of a modern country when they can't reliably use social media or understand how WiFi works? This gerontocracy is broken. Term limits and age limits need to be law—yesterday. No more hiding behind "speculation." Transparency or step aside.

I publicly and privately urged the last administration to address the public’s concerns with the former president’s health. I’m calling on Sen. McConnell to do the same and provide voters an update on his own health. Let’s end the crazy speculation. Just tell us what’s going on.

My take? I’ve been bitten most by implicit assumptions. You operate LLMs at their ideal spot. Small, modular Unix utilities, with explicit and well documented assumptions, decades of compatibility wrung out, often included in baseline training. LLMs make it seem like the are „senior engineers” with decades of field experience - and in your domain, they are. In, say, custom DSP for a custom embedded system, they are more overeager juniors, not knowing which questions to ask. Case in point. Had a system with fixed small integer PLLs between various clocks in audio domain. PDM/PCM of 60x; IO/PDM of 25x; SYS/PDM of 50x. LLM depegged it, without asking why, pushed to fractional multipliers and got super busy looping that nothing converges anymore. Junior engineering move - few months in audio DSP would teach you to check those magic numbers first. But they are magic, the ratio is not documented anywhere and we „just know” that things go bad (aliasing etc) if we somehow end up with IO/SYS that is anything but an integer. Of which 2x is the only good one on that board. The problem is, these domain specific types of knowledge exist everywhere. LLMs are pressed into service without a path for them to acquire domain specific knowledge in hard domains. Unix core utilities are in a way over-taught and over-documented compared to the rest of the industry.

This is a certain kind of talk around LLMs that I find increasingly puzzling. That is all of the people bitching that LLMs constantly generate crap code and hallucinate solutions, and are worthless for programming. This has almost never happened to me, and never during the last two model generations I have used (chat GPT 5.4 and 5.5). Occasionally a model used to get a little deranged when I pushed its context limit, but under codex that doesn't happen anymore; instead I got a red-highlighted warning when the limit has been exceeded and I need to clear my session. I've applied AI to feature changes, refactoring, and debugging over 63 different projects written in C, Go, Rust, Python, and shell. I've written documentation with it. I've decompiled a DOS binary into readable source code. It's now routine that whenever I have to touch one of my projects I start by running the regression tests, then fire up codex and asking it to audit the code for bugs and suggest improvements. My experience is that LLMs are excellent and tremendously empowering tools. Their worst limitation is a kind of architectural tunnel vision - they're extremely good at generating code to specification but sometimes blind to higher-level patterns. Which is okay, it's my meatbrain job to be good at that. The most valuable thing I find about LLMs is exactly that they *don't* screw up details and edge cases. I'm a very, very good coder by human standards (I'd better be, with 50 years of experience!) but the LLMs are better than me. Because if a code change needs to touch (say) five places in the code, they reliably find all five rather than doing the human thing of fixing four and then having to debug for hours before you figure out that there's a fifth one you missed. Are the downshouters living in a different universe than me? Are they using old, weak models? Or do they have some kind of skill issue that I can't see because I have mental habits and communication skills that are a good fit for the handles on these tools? I don't know. And I think this is an important thing to figure out, because I'm seeing lots of stories in the news that suggest billions of dollars are being wasted on misdirected token spend. It all seems very simple to me. Be clear in your thinking, tell the model what you want with precision, and good things happen. What...what am I missing here?

@Artorius519 @shai_machnes @bscholl The difference being, that in aviation you can’t decide to just fly around for a year with a dead engine. Another big optimization of hyperscalers is choosing to never replace a failed blade server - by the end of hardware upgrade cycle it’s expected that some would’ve failed.

@shai_machnes @bscholl In aviation, that turns out to be the most important reason why most airliners now have two engines instead of four.

If you’re making a new gas turbine, how big should it be? Decades ago, the computing industry figured out that large arrays of blade servers were both cheaper and more reliable than a small number of expensive mainframes. Yet the energy world is still largely powered by gigantic “frame” turbines that are hard to make and difficult to install. I think the future of energy—particularly for off-grid “behind the meter” generation —is small modular production. But how small should you go? If you go too small there is a big loss in efficiency (due to things like higher relative blade tip losses). But there’s a sweet spot in the 40-50MW range where the turbine has most of the efficiency of a gigantic unit yet is also much more manufacturable and deployable. This is the sweet spot is in total cost per unit energy, inclusive of capital expenditure. This is why Superpower sits at the ever magical 42MW.

@davepl1968 @SnazzyLabs Surprisingly I’m not a „big road tripper” but with clients all over East Coast (and the real range being closer to 200 miles than nameplate 280) - charging speed does matter. (And so does the lack of decent NACS 480V/3ph systems you can install at your industrial sites).

@SnazzyLabs Are you a big road tripper? You're always pounding the charging speed time... but with 3 EVs, I never think about it. "Overnight" is "overnight".

@EdStronski @marktilbury And it’s also a very good reminder of how much tax is there. Allowing the customer to shop around (depending on the state and purpose it might be a grey area) and move around to optimize. Price (including price of government) is a signal, hiding it deprives the market of such.

@marktilbury Re: #4 taxes, my thought - in the U.S. we have State, county and sometimes local sales taxes. If you went to a national chain store to buy an advertised $9.99 extension cord you might find a handful of slightly different prices at 6 of the company's stores in a 50 mile radius.

I'm British. Today is America's 250th birthday. Last year, I visited the USA and stayed in New York City for the first time ever. What I saw still shocks me. 7 American quirks I still can't wrap my head around:

@bscholl @JessePeltan Also because of socialism. I have some horror stories about behind the meter solar + storage at a freestanding place of worship in NYC/Manhattan (nearly ideal locations for PV+Bat, so good it actually offset the shading from other buildings).

@JessePeltan Also it is too hard to deploy new generation

The reason the mayor has to ask everybody to set their thermostat to 78° is because we don’t have real price signals in electricity. You get overcharged for 99% of the year to subsidize peak demand. This destroys the profitability of insulation, efficient HVAC, home batteries, etc. Since prices don’t reflect real costs, we get pleading and rationing instead of smart investment. Shortages are a result of the socialism that is ubiquitous in electricity. NYC is not unique in this.

@kyawzazaw @pitdesi @bscholl Time lost driving. Opportunity cost of business not undertaken. Density cost of buying properties closer to workplaces. Very, very hard to quantify, but we didn’t even start debating „ok what do we do with better cars” - less casualties, faster speed?

@qdotme @pitdesi @bscholl How is it productivity loss

Europe does some things much better than the US. One of them: Traffic enforcement In Europe: speed cameras In the US: cops hide in a bush Why waste police time on this? also: higher speed limits on highways, lower in cities