xyzzy123

I would consider asking it to model its proposed designs in openscad or build123d (ideally something query-able). Then have it render and examine plausibility / suitability from different angles. Get it to render the part in use also and give instructions to think about forces and motion.

Recommend doing this in a coding harness not a chat box.

The reason I think you might have more success with this is that the model is mostly thinking about the part in words, which it can convert to a part design in CAD in code. LLMs are really good at coding. Also means it can use relative positioning and relationships.

You will be able to iterate more easily, compare things, compute properties, commit to git etc. The process is more reproducible and steerable than generative production of images.

When the LLM can look at renders of the geometry it generated, it’s easier for it to discriminate when it’s producing nonsense like misaligned parts, things that don’t fit, etc. It’s still going to kind of suck, but it will be better. The whole process of code -> render -> inspect forces the model to put up or shut up and provides grounding. Meshes > bloviating.

As far as I know today's LLMs don't have a "visual imagination" but a process like this could be a slow approximation of one. They clearly do have SOME spatial understanding (pelican tests show us that!) but it feels really non-human.

One thing missing from this is kinesthetics. Personally I am mostly not thinking in accurate visuals in mechanical design. I am imagining how the parts feel and kind of how they move and what slips first and what bends and what feels heavy. Imagining what my hands would feel. But I don't think I trust LLMs to evaluate that stuff by writing simulation code yet.

I do feel like Oxide are qualitatively different from the others pjlmp mentioned because they're staying commodity in key interfaces like CPU ISA / OS / ecosystem (thing with big network effects) and mainly focused on fixing the control plane / management / architecture mess.

The others suffered "ecosystem collapse". With Oxide you won't be stuck on a "burning platform", your main risk is that the value prop for the hardware & management experience doesn't play out.

The other aspect is that the electrical parts of the power system are lighter at the higher frequency too. A 400hz transformer is meaningfully smaller and lighter than a 50hz one.

This is the same idea as switching power supplies but planes had to solve that before power semiconductors were cheap enough.

The power frequency also has an impact on the size of all the induction motors.

I guess if you were designing it from scratch today you would let the generator produce power at whatever frequency the mechanical engineers tell you they want to and the power electronics could convert it without problems. You would do the same thing the other way around on the motors.

You're right of course, LLMs provide a partial, unsound oracle.

The "halting problem is unsolvable" argument relies on the oracle not being able to output "not sure". But adding that option admits trivial oracles, like ones which output "not sure" for everything, so some are better than others.

The "real world" use most people have for halting oracles is as part of software safety, where if the checker outputs "not sure" you modify the software until the checker can decide if it halts.

Its funny because you can write a halting problem oracle by calling out to an LLM and have it return yes / no / not sure and get it to work reliably for almost all real code, like that is an entirely practical thing to do in 2026.

All we need now is some sort of program to evaluate halting problem oracles...

Hi, shredding the book is to reduce potential legal liability of format shifting - if you shred the book after scanning, the theory is you are not increasing the number of copies. This has come up as a factor in legal rulings. The legality of format shifting is still murky though, and the legality of training on the work are a separate question.

> Practically speaking, is there really a difference between prompting an LLM to create a CMS for a website and installing something like WordPress or Ghost?

There is some difference; when you install WordPress or Ghost, your problems are shared. There's a community that goes along with the software and a "shared understanding in the world" of how it works and where the rough edges and limitations are. A lot of the time, LLMs will actually be better at modifying WordPress or Ghost to do what you want than they will be at fixing the CMS you built last Tuesday.

When you custom build, you can get an exact fit for your needs but your misery is yours alone.

The problem he's describing comes up any time you have a team of developers. That doesn't mean the usefulness of teams is a fantasy because you didn't write all the code yourself.

If you don't have a useful mental model of the architecture that's a communication / review / documentation problem, not a problem with the entire idea of codegen.

The incident is funny on 2 levels; a) OpenAI thought so little of the model's _actual_ security capabilities that they gave it a "wet paper bag" sandbox and b) OpenAI failed to use AI to accelerate their security processes.

Food is allocated using a variety of mechanisms in peaceful first world countries, primarily money but also via government assistance, kinship, friendship, community, etc.

At any given time many people have problems with one or more of those systems. Money is easy to run out of because it's used for everything, the government can be slow and difficult, relationships can fray, people can be isolated, etc. Food banks exist as a backstop for when the regular means of allocating are not working.

The problem isn't "scarcity" per se, it's more of an allocation thing. Who has a claim on enough food to stay alive? Everyone! But what foods can they claim? How much? What specific channel / institution (with associated allocation rules) will distribute it to them? What are the conditions and controls? etc.

Allocating things can be difficult. An allocation mechanism with no controls will see fraud, waste and abuse. Even when an institution is willing to give things away no questions asked, there are (often invisible until you think about them) conditions like "please don't claim huge quantities and resell what we're giving you, that would be unfair to others".

It's also interesting to think about the fact that you can't fix food scarcity in general by simply giving hungry people money, because money is too fungible.

I like your moxie and I also agree Australia is an over-regulated racket... but I'm also glad my neighbours aren't mass farming frogs or cockroaches in my suburb without talking to the council first.

My brother had a neighbour build a crematorium next door in NZ; not an industrial or rural area, just the suburbs, no permits, no consultation, no notification. Very entrepreneurial. They got shut down but the flip side of "you can just do things" is surprises like a plume of smoke and the smell of burning corpses when you step outside.

If we take a contested issue like the origins of COVID-19, the facts are detailed, complex, contradictory and ambiguous. They also don't make narrative sense without some degree of interpretation.

Interpretations of complex webs of facts often diverge along tribal lines and bias can easily leak through when editors need to decide which interpretations are authoritative.

In Sydney Australia its < 2 days of median rent.

If you jerk off a horse just for the love of the game, you're a criminal and that's abuse. But if you're paid to do it (e.g. for insemination) that's fine. The act is the same, seems doubtful the horse is harmed. What society has a problem with is the fact that you enjoyed it.

For some reason with people it goes the other way around.

I feel like the build threshold discussed is _extremely_ optimistic?

> But does that always hold true? Let’s take the other side for a second by examining a much higher-priced SaaS product. Gemini reports that the price of a fully loaded Salesforce seat is ~$500/mo. Say you need 50 seats, that’s $25k/mo!

> For that price you could have 1.5x engineering resources (25 / 16.7) working on your clone full time. Once again, a CRM’s a reasonably complex piece of software and a rebuild wouldn’t be trivial, but no matter how you construe it, this is closer to a “build” decision, even for a smaller company. (And with Salesforce down 30% YTD, the markets seem to believe it too.)

If $25k for salesforce is too much, my view is that your first thought should be to look for a cheaper competitor, not build a thing?

Ok, you vibed your own. Great! Now you need custom integrations for everything, you can't hire out of the market, you have to re-train everyone and all new starters, you have an extra thing to HA, monitor, back up etc. People can't google for answers about it anymore. This is before we can talk about what it actually costs in terms of bikeshedding, roadmap creation, project management, product management etc etc. Plus compliance, security, your org policies and relevant regulations if you're storing personal information. Think of how many meetings it would take to get this done, the political costs, and how much it costs to get consensus in big companies.

There is also RISK. Nobody is gonna get fired for choosing SalesForce, but there are many different angles by which building an in-house solution for something considered commodity (tho an expensive one) can go horribly wrong.

The more subtle cost is _brain space_. Human engineers have context limits too; switching projects has costs, and you can't put ONE engineer on the thing and expect it to be sustainable. You need about a minimum of four people to understand anything you expect to maintain and operate long term.

Your org's capacity for engineering focus is precious and IMHO you should try really hard not to use it on non-core stuff unless you have to.

I wonder if what you're noticing in Fabrice's code is a lack of _abstraction_ beyond whats obviously needed to get the job done. It's not spaghetti IMHO, I think its what code looks like when you're smart enough to just hold most of the problem in your head. I am speculating a bit here, because I am not that smart.

If I had to describe it in aesthetic terms I would maybe say brutalism?

I use hermes with crons and kanban for this. I have a "self driving" homelab where upgrades & config changes happen in a non-prod ct before being applied to "prod" (just, a diff ct). Its mainly compose stacks with data volumes on zfs that gets snapshotted a lot to limit possible data loss. I use the orchestrator to break down regular tasks like "health check the whole env" or "find all containers that need an update", "update terraform providers" into changes with their own card & MR, actioned serially via kanban. The task detail is just in playbooks in the repo, like a human would use. Executed by qwen 3.6 27B. I still use claude to help define and oversee workflows though.

Probably CI is the better primitive, I usually use gitlab though and I felt "put off" by all their native AI features being license gated.

> You and I can still trade with each other, no robots need be involved.

Unless one of us happens to be a food producer we will both starve. We need our trade graph to be connected to resources we need.

Production also tends to need exclusive access to resources (land, materials, etc) and you will be competing with machines for access to those.

> The reason we have an economy and money and trade is that we need to incentivize people to produce all the stuff that people consume, and manage those finite resources constrained by people’s finite time. But you can do away with all that messiness of all that exchange and just have AI micromanage the economy. AI should be able to figure out how much to produce, how to limit waste, who should get what, etc. in a very fair and efficient manner.

Who owns the robots though (plus scarce exclusionary inputs), and how are you connected to the part of the trade graph that produces abundance?

> If there’s no limit on production, and no need for human labor, then we don’t need to incentivize people to work, or try to bound the amount people consume by the value of what they’ve produced.

This is very much a question about who controls the means of production.

The materials in a 60hz transformer I think are the main BOM cost for linears? Relatively, transformers are say 2x cheaper than in 1970 (steel and copper are relatively MORE expensive than they used to be but the manufacturing improved) - but chips are more like 100x or 1000x cheaper. The high frequencies of switchers let you shrink the transformer (less materials) so its a big win.

Also the waste heat of linears constrains your design in terms of weight, power density, how big you need to make the enclosure, etc.

For reference I looked at what I'd need to buy on AliExpress to power my laptop off the cheapest linear PSU I could find (say 20v 5A), something like: https://www.aliexpress.com/item/1005003397909174.html - it weighs 4kg vs 300g for the charger I have. Apparently there ARE audio people who put their NAS on a linear PSU (I did not realise that was a thing!) but I'm going to try to forget I saw that.

The real sandbox is not caring if your computer gets bricked.

Recursive improvement for me but not for thee.

Right; most kids won't be very interested unless they're around other people who seem genuinely interested. They key their learning effort off perceived relevance. Perversely, bad school environments can kill interest instead of nurturing it.

The very best environments will have other kids who are engaged. It is induction into a culture, with its own language, ways of thinking and values. If you can't find that directly, you have to do your best to create a mini version as a family.

You can turn almost anything into a little maths games, card or dice games provide an opportunity for probability puzzles, the environment provides things to estimate and count, you can challenge them on car trips with rates and distances, etc etc etc. How many sides does a circle have?

Feels like an interesting trend where a "solution" for parent anxiety (and to be fair, vastly increased societal expectations around what "care" looks like) is proposed to be electronic surveillance.

It's a kid tracker / ankle bracelet in an attractive form factor.

I was a kid in the 80s, city fringe, single parent who worked until 5:30. Honestly nobody had any idea where me & my friends were a lot of the time. Totally acceptable in that era.

The main worry I have about tech like this is, at what saturation of deployment does the norm shift such that it's irresponsible NOT to electronically track your kids whenever they leave the house?

It is not "multiculturalism" in the way that word would be used in say, Canada.

They have quotas to prevent enclaves, they actively manage immigration to keep it about 3/4 Chinese, there are a lot of restrictions on speech and the ways you're allowed to organise.

The things Singapore does to manage their ethnic diversity 100% would not fly in the west.

I feel like the shape of the market right now for "home lab" inference is:

The sparks are good if your ultimate plan is to spend even more on NVidia hardware in future to run your dev setups at usable speeds. Or, you're developing for a work cluster.

If you mainly want to run local models at acceptable speeds portably, buy a mac with lots of RAM. If you’re happy with non-portable / racked, buy 3090s (dense) or mac studios (MoEs). Buy newer cards if you are restricted on power or slots. If you are rich, buy a6000 blackwells.

Or like a m4 max? This thing has <300GB/s vs the max with 550GB/s

All those CUDA cores in the sparks but they're starved for memory bandwidth.

I am still waiting for NVidia to release a system that legit beats 3090 maxxing for the home gamer...

From the perspective of pre-history, anyone living in a western country is in a post-scarcity society.

Practically no one starves. The murder rate is down by crazy numbers. I'm not sure how much of a problem racism was at that time, but you wouldn't have had many chances to meet people from different haplogroups in any case.

The abundance created by industrial society is not distributed "evenly" or "fairly" but the baseline shift is insane.

Thanks; I didn't spot that they disabled tools in the harness. Also they don't provide an "out" to allow the models to express uncertainty so the instructions force a guess to be made.

As an aside though it's still funny that the two tools WITH search also disagreed.

> "On May 18, 2026, Ukraine carried out a drone attack on Moscow, Russia"

I actually don't know which way you came down on that one?

I think strictly it's false but "mostly true" would be justifiable? (as in, to say it's false would be misleading if it lead the reader to assume there was no attack around that time).

https://www.washingtonpost.com/world/2026/05/17/ukrainian-dr...

It seems it happened Saturday 16th overnight into the 17th, not the 18th. I see this a LOT with fact checking. It shouldn't be this way, but political bias seems to nudge people into making calls land one way or the other with selective application of pedantry.

iOS always had a strong focus on security but if you take the time period say 2005 - 2015 it did not seem like there was much investment in macOS security at Apple. I am talking about stuff like exploit mitigations and relatively low hanging LPEs. Features like (full) ASLR / SIP / kext controls were added well after competitors.