Claude will load the name and description of each enabled skill into context at startup[0]; the LLM needs to know what it can invoke, after all. It's negligible for a few skills, but a hundred skills will likely have some impact, e.g. deemphasizing other skills by adding noise.
DIYers and enthusiasts should still worry about their house burning down because one of these boards started a fire. An insurance company would investigate and find any excuse they can to deny payment.
Insurance covers the insured's own gross negligence. This is a trope up there with believing the "warranty void if removed" stickers.
The real reason this should give you pause is that you don't want your house to burn down regardless of an insurance payout. That is how your incentives remain aligned with the insurance company.
I’m not sure about your home insurer, but here in the UK, most home electrical wiring, including solar for example, is “notifiable” to building control and requires a qualified tradesperson to implement.
If you want to DIY everything you need self-build insurance.
I’m definitely not covered if I were to burn my house down with a dodgy inverter installation.
In many (most? all?) jurisdictions in the US there's a carveout to exempt DIY work from licensing requirements and to require insurance to cover it.
Many things you still have to get a permit and have it inspected afterwards just like a professional would. But if you skip that no one will ever notice unless it's major structural work (such as a new deck).
That said, if it's electrical and you skip the required inspection and then your house burns down and your work was at fault that might nullify your insurance but I'm not sure.
IANAL but I don't think that's true at all. Electrical work needs to comply with code standards (NEC). I'll eat my hat if an insurance company pays for a fire that was caused by using homebrew electrical systems.
I'd really like to see citations to the opposite. After all, the main function of most homeowners' insurance policies is to keep banks whole regardless what happens.
In my experience insurance companies just ask you a bunch of big-picture or specific questions (eg swimming pool), reserve the right to inspect even though they never do, and then jack the rates rather than trying to grok more unknowns.
I've never read an insurance policy where anything like this is explicitly mentioned. I suppose there is a legal path of being criminally charged by fire investigators for having performed unsafe wiring (non-Listed power handling equipment as part of the fixed wiring), and then the insurance company denying you because of that criminality. Or in states where DIY wiring is not illegal, perhaps declaring that the wiring itself is "illegal" (as it goes against the NEC (despite the NEC not being openly published as we generally expect from laws!)) and then hanging their hat on illegality regardless of criminality? But does any of this happen in practice?
Surely if insurance companies were concerned about fires caused by dodgy electronics, they'd address all of the people using non-NRTL-tested GENSYM brands from Amazon et al? They've got no ability to post-facto deny based on this (said devices aren't illegal), but they could surely make it an explicit condition of policies.
Like I said, I'm not a lawyer.. and I'm no longer a homeowner so I don't have an insurance policy. So I don't have a citation for you. I may be wrong about this, but I don't think that I am. If anyone who's in the know can help then I'd be okay with being proven wrong.
There are certainly cases where harmonic distortion is a problem for a device. It’s just that everyone is left guessing, and there’s an overblown fear of devices being harmed.
Sure. It’s just rarely what lay people associate with “sensitive”. Most customers are worried about small electronics with switching mode power supplies that wouldn’t have a problem with just about any power source.
I wouldn’t run some AC motors, old AC clock, ham radio, or many other things on some generators.
The line is open to interoperation and never defined by the manufacturer. It’s blanket liability avoidance that confuses customers.
Should is doing a lot of work there. The reality is most don’t. These are people who don’t understand minimum conductor size trying to DIY a solar system.
So in the UK you won’t get sign off from building control if your electrician uses a dodgy uncertified device, and that results in your insurance being invalidated, at least according to my policy.
The cheap stuff usually have the certification labels printed on them, but a cursory look inside usually shows breaches if you know what to look for (like we have had stuff from China where the ground wire wasn't connected to anything, but even if that's right there are things like just not having enough clearance between mains and isolated DC in power supplies etc.). So they are definitely not actually certified, they just slap the labels on anyway.
Of course, there is a rising group of actually really high-quality native brands starting to come out of China - Anker were an early one of these, Xiaomi is a massive one, UGreen is really popular on Amazon etc. These kind of brands actually will be properly certified and are scary good for how cheap they are. So it's changing a lot.
That's basically what most software Open Sorce projects state in their licenses as well: Use at your own risk. As someone who publishes design files openly you really don't want to be liable if someone uses the hardware in a wrong way.
That being said, the Libre Solar components are also meant to be used as the basis for customization (hence, called building blocks). Some of the devices are used with minor modifications in certified commercial products.
Apart from boats/caravans, DC systems are used a lot for rural electrification in the global south. This is also where the communication features of Zephyr RTOS are very important.
this is a really good point that I neglected in my original dismissal of the project: it's going to be super useful for DIYers and small shops in the less regulated parts of the world.
Not that this is accepted by insurers or AHJs ("authorities having jurisdiction"), but one can use UL-certified components in an (open-source) _assembly_ that itself isn't UL certified. This at least supports the argument that the overall product is safe if thoughtfully designed and assembled. An example is the OpenEVSE level-2 car charger (which I had a really good experience with).
There's nothing stopping a company from creating products based on these, and having them certified, assuming they satisfy the requirements for certification (which if they don't you probably don't want to put it in your house anyway). I'm not familiar with the cern hardware license, but it appears to allow commercialization.
If you're not selling them then there isn't really any certification requirements until you start to get into the mains voltages. This all seems to be 12/24V DC stuff on the battery side and 60V DC on the solar side which comes under 'extra-low voltage' in basically every jurisdiction which is usually not really regulated because it's difficult to shock someone.
The battery stuff is more risky (bringing lithium cells into the picture) but I don't think anyone should be worried by the MPPTs.
I don't know much about UL but I can say that FCC certification (also technically required) for electronics can range from about $3k to something like $30k depending on what you're doing.
($3k would be for "unintentional radiator" device, i.e., not supposed to be a radio, $30k would be for "intentional radiator" device, i.e., supposed to be a radio)
FCC ensures a product doesn't cause radio interference, while UL ensures the product is safe to use and won't cause fires or electric shocks. For DIY, your primary concern is UL certification.
Because of customs product import rules, that FCC stamp is often not optional. Now if it was a dodgy seller, the stamp will not match the physical devices on rare occasion. =3
That roughly lines up with what we paid* to get CE and safety stuff done for a small battery-powered product with a radio on the EU market (primarily in the UK).
*Testing and tweaking and then sign-off in grown-up labs.
In general, almost all insurance companies will demand UL stamped hardware, and most mortgage/lease/commercial property requires insurance.
A hobby BMS is usually a bad idea, as most kits from unknown origins prioritized cost over safety. Depending where you live, prior to roof installation there may be additional zoning and signed engineering drawing requirements.
It is not hard to find UL equipment, but expect to pay about another $600 for the BMS. Yet, it is better than a house burning down, and the insurance provider denying coverage.
Have a look at local certified installer companies, and make sure to get some real references in your town. Just like most HVAC companies... some installers are just over priced scams. Some folks claim https://www.pegasussolar.com/ was inexpensive, and might be worth a call. Best regards =3
It doesn't have to be a certified install company in many places, it just needs to be inspected which most states will do for homeowners. (if your area is different contact your representative)
UL or other certification is a very good idea. They can't automatically deny coverage for lack of certification, but it becomes a much harder fight for you to prove the non-certified equipment wasn't at fault.
I wrote it up. The full system reference is here:
https://blakecrosley.com/guides/obsidian — vault architecture, hybrid retrieval (Model2Vec + FTS5 + RRF), MCP integration, incremental indexing, operational patterns. Covers everything from a 200-file vault to the 16,000-file setup I run.
I implemented this as well successfully. Re structured data i transformed it from JSON into more "natural language". Also ended up using MiniLM-L6-v2. Will post GitHub link when i have packaged it independently (currently in main app code, want to extract into independent micro-service)
You wrote:
>A search for “review configuration” matches every JSON file with a review key.
Its good point, not sure how to de-rank the keys or to encode the "commonness" of those words
IDF handles most of it. In BM25, inverse document frequency naturally down-weights terms that appear in every document, so JSON keys like "id", "status", "type" that show up in every chunk get low IDF scores automatically. The rare, meaningful keys still rank.
For the remaining noise, I chunk the flattened key-paths separately from the values. The key-path goes into a metadata field that BM25 indexes but with lower weight. The value goes into the main content field. So a search for "review configuration" matches on the value side, not because "configuration" appeared as a JSON key in 500 files.
MiniLM-L6-v2 is solid. I went with Model2Vec (potion-base-8M) for the speed tradeoff. 50-500x faster on CPU, 89% of MiniLM quality on MTEB. For a microservice where you're embedding on every request, the latency difference matters more than the quality gap.
False. SpaceX development of Starship is much cheaper then SLS despite using more test vehicles. The claim that building hardware rich is more expensive is not really shown in the data.
NASA has done some analysis on early SpaceX and shown that their methods produced a 10x improvement in cost. And that was with the method NASA uses that often turn out to be wrong.
At least there's a good reason there - they're easier to clean. That's not much of a concern with microwave controls.
But I disagree with the idea that we don't need precise times on a microwave. The article / book disagrees with that, and the think I most regularly microwave (milk for my kids) needs 1 minute 50 seconds. 2 minutes and they'll reliably complain it's too hot.
The real problem with microwave UX is that the interfaces are often simply bad. People think the power/time dial interface is good but that's because it's difficult to mess it up (though they usually manage anyway by having them go up to 30 minutes or whatever).
It's really easy to mess up a button interface but you can also do it well. My microwave is close to doing it really well. You press a high/med/low button, then 1s/10s/1m/10m buttons to the desired time, then start. The only things they got wrong are that it requires pressing the power when 99% of the time you want high, and you could probably get a more useful distribution of time increments (I'm literally never going to use the 10m button).
But apart from that it's nicer than dials, which are often very cheap and imprecise.
I've never had an issue cleaning the dials. They're smooth hard plastic, and they don't get particularly dirty.
>though they usually manage [to mess up the interface] anyway by having them go up to 30 minutes or whatever
What's the issue? I've microwaved that long before.
>My microwave is close to doing it really well. You press a high/med/low button, then 1s/10s/1m/10m buttons to the desired time, then start.
We're very different people! That UX sounds dreadful to me, one of the worst I've heard (and unfortunately encountered).
Enter time on the keypad, optionally press Power and enter that, press Start. Also needs a Plus 30s button. This is the one and only correct way to implement a push button microwave. ;)
I count five presses instead of 3 to get 90 seconds, including one way that's just pressing the same button 3 times (+30s).
>needs 1 minute 50 seconds. 2 minutes and they'll reliably complain it's too hot.
Seven presses?
The dial microwave I use can distinguish between those two. It helps that the shorter times are given more room, so you can adjust them more precisely. 1:50 vs 2:00 will make a difference in my experience, but 7:50 vs 8:00 generally won't.
You could have a hybrid approach of course, but then I suspect the engineering tendency would be to "lock in" the time after starting the oven, so it can't "accidentally" be changed.
Looking for a photo of my microwave dial, I came across this surprisingly relevant post:
> What's the issue? I've microwaved that long before.
Really? What for? Anyway the vast majority of microwaving is going to be in the 1-5 minute range. By making the dial linear and giving it a huge range up to 30 minutes, you end up making e.g. 30 seconds and 1 minute impossibly close.
The commercial microwave oven someone else linked had a solution - make it logarithmic.
> Seven presses?
Eight actually, but it really is quicker and easier than doing the same with a dial though. I agree it could be optimised though. It shouldn't be necessary to select the power and a 30s button would be good (down to 5 presses).
The vast majority of microwaving is in the 1 to 5 minute range only for those who use a microwave oven only for reheating.
For cooking, times from 10 to 15 minutes are more frequent, though things like potatoes or sweet potatoes need only 7 to 8 minutes. Only a few delicate vegetables or fruits may be cooked in the 1 to 5 minute range, e.g. onion, garlic, leek, parsley and dill, etc. Meat needs to be cooked at low power, which in turn requires long times, typically over 20 minutes. There is also a very small number of vegetables that need cooking times over 15 minutes, e.g. the common beans, for which even times of 30 minutes may be needed.
That said, all the microwave ovens that I have used (in Europe) had rotary knobs with variable resolution, fine for short times and coarse for long times.
For many years, I have also belonged to "most people" and I was cooking in one of the weekend days by traditional means, which required many hours, then in the rest of the week days I was reheating the food in a microwave oven.
Only a few years ago I began to experiment with cooking raw ingredients in the microwave oven. After discovering how much this simplifies cooking I regretted very much that I had not tried to do that earlier.
Because cooking at microwaves is much faster, nowadays I cook most food immediately before eating it.
I have used only microwave ovens with rotary knobs.
They had a finer resolution of 10 seconds for short times, then the resolution was progressively coarser for longer times, e.g. of 30 seconds for times over 10 minutes.
This is perfectly adequate for finding optimum times, and I cook in a microwave oven all the food that I am eating, from raw ingredients.
I don’t disagree with your point about agent abilities with older, concise, well-represented-in-training-data tools though.
reply