throwaway219450

Seems about the only kitchen things you can trust now are cast iron, carbon steel and sheet pans, because you can’t really do much to cheapen them in terms of materials except take them offshore (and most of the brands like Lodge are proudly domestic). After spending some time in culinary I started buying stuff from restaurant stores. It’s not always pretty but it tends to be a lot more practical. Glad Cambro is on there.

I’ve also started to pick up ceramics from local makers as there are a lot of potters lurking around. Often the prices are good, certainly comparable to the luxury consumer brands like Wedgewood.

Also we’re a Makita household, not because the usual options are insufficient for home use, but they’re one of the only independent power tool makers left. Most other hand tools I try and buy vintage/restored when I’m in Europe as they’re much cheaper than in the US.

Point6 as a successor to IceBreaker I didn’t know about. There are very few good merino companies and apparently their conglomerate also owns Smartwool.

There isn’t much need in the UK. Farmers use tractors or vehicles like the classic 4x4 Land Rover, Subarus, etc. None of the roads or the infrastructure are designed for American size pickups, especially not the countryside. If people need to move something big, they hire a van.

People are also extremely sensitive to fuel costs and an enormous car that guzzles gas and can’t be parked on a tiny village road is a downside. Bear in mind gas in Europe is like $8/gal and salaries are much lower.

The statement car of choice for people who don’t need the functionality is the modern luxury Land/Range rover (ie the Chelsea tractor).

Nevertheless this PoW is used by some conference venues now, in the case of a dispute against human authorship. NeurIPS does allow LLM use for "minor" polish like spelling, grammar, correcting non-native speakers. But you can't have the AI write the entire paper:

See https://blog.neurips.cc/2026/06/02/ai-generated-papers-in-th...

Unfortunately, given the volume of submissions that appear non-compliant, relying on author declarations is insufficient.

...

Authors whose submissions show significant AI involvement must provide an audit trail that clearly demonstrates that they complied with the policy. We expect that in future years this kind of audit trail will become a default.

The method uses Pangram, but that's somewhat arbitrary. The authors must provide the draft that was used to prompt the AI and show how the changes were made, which should be possible with a chat history. Where it doesn't work is if the authors claim they did not use AI, but otherwise it's quite hard to spoof. Edit history on its own isn't good enough because you could trivially have a program emulate keystrokes to type at a human speed into a document with tracked changes; the proof here accepts that AI was probably used and aims to determine to what extent.

I do think this goes back to if the text is high quality enough that humans don't notice, is that OK? Probably yes. What we don't want is poor quality writing from any source.

This is also how Arduino emulates analog output, since most micros don’t have a hardware DAC

https://docs.arduino.cc/language-reference/en/functions/anal...

Arguably these are different use cases. Cheap fast-charge cables are usually limited to full speed only (linked ones included). For that reason they don’t meet my criteria, as you can’t interchange the charging cable for a data transfer cable.

I’m not advocating using USB4 cable for everything, but I also don’t want to bother testing cables when I can buy a reliably good one for twenty bucks. I’m sure a cheaper equivalent exists.

Cost, really? A 2m USB4 cable from Belkin is $20 (eg INZ004bt2MBK). You don't save much going to 4 ft, sadly.

I find their cables to be pretty good (the TB5/USB4). They're certainly cheaper than Apple's. They have a soft rubber coating (not quite silicone) and a good bend radius. Compare to the super-stiff (but good) cables that Dell provide with their monitors.

I had a discussion with folks at work about what information is worth retaining in the face of AI doing everything for us. A lot of what we have in our heads to qualify as "domain experts" is pretty esoteric. How to invoke command line tools, gotchas because library A uses one convention over library B, AWS vs GCP; so much is specific to a tool rather than a method. There are also a lot of entrenched tools that are effectively unfixable due to the risk of breaking changes, so you have to shrug and accept + learn that's how it works.

Catch-22 is it's still important to know the fundamentals so you know what to ask for, but if you don't know the esoterica, the model is eventually going to make an assumption and screw things up. And the models don't have much taste either in prose, or in coding/comment style.

The UK has de minimis for a while yet, and PCBs are 0% import duty. The only thing you might get shafted by is Royal Mail's handling fee if it gets checked by customs. We're talking about hobbyist PCBs here, not production runs.

If you write Qt applications, Qt Creator is perfectly usable and as it's cross-platform you only need to be familiar with one IDE. It ships with its own deploy command which packages the app. You still need to run code signing/notarization last time I checked, but if the complaint is the XCode GUI then no problem there.

In the past if you didn't live in the US, OSHPark was a great option. Local fabs were very expensive, had big setup fees and even the "maker" places charged more and you'd only get HASL. Meanwhile OSH gave you 3 with ENIG and the quality has always been excellent. They nailed the automation to make group buys cost effective.

If you live in the US, it's free shipping and for breakout boards or smaller work, still a great option. Unless you're a student, I think most side projects aren't make or break if the boards cost $5 vs $2. Even $20-25 I would consider fair for a one-off side project and they also do $1/sqin medium runs if you need > 10.

This is a fascinating thought problem.

On the extreme end you have recipe blogspam, with the tropey life stories and Amazon affiliate spam that precede the meat (pardon the pun). If that pre-filler is inconsequential, why not have an LLM parse the semantic markup - which 99% of recipe sites use for Google SEO - to produce custom content for you? Like a history of the ingredients, or better instructions with an equipment list tailored to your kitchen and timings/temps for your specific oven?

I can see there being a middle ground where you publish some sort of context that a user’s LLM can generate the content from. Is it fundamentally any different from dramatized non-fiction? Think of books like Operation Mincemeat where the facts are thoroughly researched, but the author uses a lot of artistic license to tell the story.

But to your point, if there is back-and-forth, I wish people would pay more attention to the obvious traits of LLM writing. Not everything needs to read like it’s a snappy editorial, but I sense we’re still in the early days where people have been handed this technology and are simply excited that they can pump out 1000 words in a coherent narrative.

Oh and of course the $64k question is this: if an AI generated article is indistinguishable from a human written article and it is accurate and interesting, do you care who wrote it? We want to avoid low quality, not AI generation, right?

Right now, in mid 2026, many of believe that LLM prose is not as good as human output.

The deception is one aspect. AI for spellcheck and proofreading is one thing; AI writing the content wholesale in the first person, which you pass off as your own, is another. I also don’t care for the Tailwind-styled demo sites that have flooded Show HN.

The sort of content that people react viscerally to (as slop) is rarely that subtle. Again, no problem with careful and subtle use (see CGI in movies), but when I see Claude-isms and snappy section headings that nobody used to use, I close the page.

You can/should write for you, to get experience writing, to present a portfolio, to share things you find interesting, a photo a day, whatever. Why does that have to die?

I like doing this if I can’t sleep, although there have been occasions where I ended up staying awake to finish the whole thing.

I don’t read in bed unless I’m on my own, or we’re both reading, as I’ve not found any satisfying book lights and I don’t use an e-reader. Also probably better for sleep hygiene and, as I get older, ergonomics to have a cosy spot somewhere else. Younger me could read folded in half, older me doesn’t want the back trouble.

I haven’t tried, but it looks like you could trivially solve with optical flow?

Edit: looks like yes, from the shared chats people are posting. But it’s interesting to think of communication schemes that require a temporal component so any single image is unreadable and can’t be beaten by long exposures or other tricks (otherwise persistence of vision displays would satisfy). A sort of physical anti copy/paste.

The actual prose is ham fisted, but the structural bits - “why it matters” headings and bulleted content are traditional clickbait. Compare to “the results may surprise you” and similar stock phrases that we roll our eyes at. This stuff has always hooked people so it’s unsurprising that RLHF got its way.

If you want examples look at popular DIY magazines from a decade ago, they’re full of this sort of material where every article subheading has a catchy sentence.

I’ve found Codex’s overage to be much better value than Claude’s. A monthly $10 budget is plenty for my backup Codex usage, but on Claude Code that would be gone in a couple of days.

A lot of human PCB errors could be caught by analyzing the netlist against requirements and knowledge of the datasheet. Schematic and board formats are usually plaintext, so you can generate those directly even. Or kicad + python?

That’s probably enough for an LLM to check if you’ve mis-wired something, missed a part or chose the wrong resistor to set a regulator voltage. Plus good old DRC/ERC. If you pass all of those, there’s a good chance things will work unless your placement is really bad, but you could manually lay out and autoroute a lot of simple boards. Not to belittle the parent but a 4 layer board is actually simpler in some ways because you have a power plane which is one less net to worry about.

For analog work you can even run a SPICE simulation.

Going to need more than that. Why? Do you want to work in biostats or biotech, or do you enjoy it enough that the money and time commitment is worth it purely out of interest? What skills are you missing that you think a masters would give you? Online, in person, where?

If you’re cooling with whole house AC, you’re probably fine if there’s a register in your bedroom and the air is free to return.

Right, but that also means sourcing a grouphead, a suitable boiler and all the other bits. Plus worrying about thermal cutout, running the boiler dry, OPVs, all sorts of safety things. Having someone else handle that is nice. Those projects also exist, but the point is you can buy a platform that has all of that in a nice chassis and only need to replace the control system by mitm-ing the wiring.

It looks like a cool project for sure https://www.diypresso.com and actually reasonably priced for what you get and what a similar spec system would cost off the shelf.

The biggest benefit for me has been getting side projects done. Little bugs or todos that would take more time than I have to spare, but don’t detract from the project.

I’ve had Claude do stuff like interrogate IoT devices, look into proprietary file formats and even assisted debugging electronics from a text description and datasheets. It has stupidly broad knowledge as long as you watch out for things like getting part numbers/boards correct. Have you ever tried to ask for help on an electronics forum full of miserable greybeards? $20/mo for getting ballpark answers immediately is excellent value.

Do assess if you want the learning experience or not, and whether you care if the problem is solved for you.

For writing I get the appeal to bloggers, but I can’t stand reading anything that has the hallmarks of generative AI.

Gagguino is a great example of this approach, licensing scandal aside†. Espresso machines are expensive and not because the software is particularly clever. They are electrically simple, but mechanically there's a lot of pressurized plumbing that you really don't want to DIY.

The kit is a control board for the pump and boiler, and some add-on sensors for temperature and pressure. The "high end" features that it enables are almost entirely software driven, the main one being temperature control via PID. I've seen even simpler mods for other machines that bypass the "brew" button so you can do things like connect a bluetooth scale to enable brew-by-weight on a machine that doesn't support it, or add a shot timer.

The commercial version of this would be the Decent, but it costs 3x as much. I would love something like this for my robot vacuums. Valetudo is minimally invasive, but there's no reason you couldn't control the vacuum + wheels, but navigation is hard and those sensors are much more complex (can you even access the camera and undistort the image?)

https://gaggiuino.github.io/#/

† they pulled the rug on open firmware

Being circular allows for zero clearance turns.

If you imagine a square robot traversing a wall and approaching a 90 degree inside corner, it can’t make the turn and would also be unable to make a perpendicular move to get more space.

That said Eufy, Makita and others make square-ish robots.

We do this with movie night now. It can be 15 bucks to rent an HD movie - not even a new release! Frequently it’s cheaper to buy a copy and give it to a thrift store afterwards.

I’m considering trying one of the mail order rental companies again.

The issue I have is the documentation and “status” is slop. Looking at the repo, how much of it is even real?

There’s supposed to be a build-along on YouTube but nothing there yet. The BoM is a bunch of aliexpress modules which is ok, but what about the chassis? Is that image generated?

The RFC calls to generate accurate models for the components, but the render looks like a full assembly?

Try to avoid milling unless you absolutely need it. Better to go 2D with some tolerance and print small adapters, or use standard T-slot hardware, to connect it. It's often educational to browser McMaster just to get a feel for what standard parts even exist.

For a bracket I would look at a laser or waterjet service. Sendcutsend is one of the more well known web shops and the pricing is OK for the convenience. Also look out for local places that are linked to education. Some libraries even have laser cutters. One shop near us will do simple jobs for machine time + material cost, like you give them a DXF and they'll cut acrylic for a good price. Other option for metalwork is to join a makerspace. The dues are often very reasonable vs trying to get a lathe into your house.

It's not obvious from the marketing if this is applicable to non-biosciences. If it is, they couldn't come up with a single example from another domain like astrophysics?

https://www.anthropic.com/news/claude-science-ai-workbench

EDIT: Installed the app, it has zero connectors for non biology, which is a shame. I assume they'll come later.

Same advice as ever? We call it context engineering now, but prompt engineering still matters a lot. Most of the failures I run into are unspecified assumptions made by the model that derails the conversation, but usually updating the first prompt fixes it. Opus in my experience is a bit better about checking assumptions, while Sonnet will plow on ahead. An example is mentioning a file that doesn't exist: Sonnet will go ahead and try to grep your entire hard drive for it. Opus will say it's not local and request the path.

I trust neither for general knowledge and I still find Opus giving me answers that are completely BS. But the token spend for Q&A is nothing compared to coding, so I always use Opus + a lot of thinking. For coding, I find Opus to be better value/token but I haven't done any sort of rigorous test.

Intel was supposed to build a fab in Magdeburg, which would have been great, but apparently the reason it was canned (2025) was they couldn't secure enough customers.

Well “humane” and fair aren’t necessarily the same, and some people hate loops.

I like programming problems, spending a day at Google was fun, they put me up in a fancy hotel, and the interviewers were nice. Like it was clear a lot of time and money had gone into the process (6-8 hours of dev time is not cheap), not a zoom and ghost like most companies.