briHass

An old PC ATX power supply is great for most things that use 5/12V. You get the 2 most common voltages for network gear, with plenty of capacity and efficiency (with a decent PSU.)

You do have to solder, though I wouldn't be surprised if one can find sata power to barrel adapters on AliExpress.

It's the other way around, unfortunately. The senior engineers will still be useful for architecture and infrastructure considerations, as well as guiding the agents. It's the junior engineers that get nailed, because there's little incentive to hire one when a LLM does a better job immediately and costs less.

GitHub Copilot was the only one with absolutely insane subsidies, where they metered by 'request' instead of tokens. A request that costs 3 cents could end up burning $20 worth of tokens or more. That ends this month.

I was actually quite worried, because I've been using GHCP for large chunks of work, but the billing estimator they released shows I was only at about $150-200 a month in API priced tokens. Sure, that's a subsidy for my $20 subscription, but not insane.

Heavy use of agentic coding tools, in a responsible manner, probably lands somewhere around that $200/m mark at API pricing. Assuming that makes the provider money, I don't see that being hard to swallow for businesses employing developers in Western countries, given the hours it can save.

The real risk here is to personal project vibe coders. Building a huge app by abusing subsidized plans is ending.

If their pricing turns out to be what they claim, and copilot cli has accurate token counts, they had the best deal around.

Just today, when I wasn't being especially chatty with GHCP, I used about 12 requests to get a few thousand line changes in 3 projects I'm juggling. The last project repo of copilot I closed, in 3 hours burned 38M input tokens, 28M cache, and like 400K out. For GPT5.4, high. That's like $135, in half the day, 1 of 3 instances. No crazy tool use, just lots of docs and unorganized code. GHCP charged like 70 cents for that on the old plan.

I'm more optimistic about LLMs tracking down and fixing issues in software, even without SO/forum posts, at least for OSS. I've seen enough unique insights from agents on tricky problems to know it wasn't extrapolating from a helpful comment somewhere.

It hit me that as it's deciphering some verbose log file, it has also read through all the source code that wrote that log, and likely all of the discussions/commits that went into building that (broken) feature.

Have you tried the Import plugin in Obsidian? I used it the other day, and it seemed to do a good job with OneNote, though my ON notebooks are just simple text and images. I believe it automates an export from ON to HTML and then converts that to MD.

Once they're in MD format, you can always get your favorite agent to do more formatting and organization, which was one of the big reasons I moved away from ON.

Nonsense. WinGet has the ability to add repositories, just like any other package manager. If you want the 'approved' packages for the distro, that would be the msstore repository. If you want to use the 'community feed', which WinGet warns you about the first time you use it, it's less vetted, but still goes through Defender scans and community moderators.

If you go adding any old repo to APT, you have the same risk. You should look at how much code review goes into packages for major distros like Debian, hint, not much, especially once the initial package was accepted.

Disagree entirely.

GHCP at least is transparent about the pricing: hit enter on a prompt= one request. CC/Codex use some opaque quota scheme, where you never really know if a request will be 1,2,10% of your hourly max, let alone weekly max.

I've never seen much difference with context ostensibly being shorter in GHCP, all of the models (in any provider) lose the thread well before their window is full, and it seems that aggressive autocompaction is a pretty standard way to help with that, and CC/Codex do it frequently.

Windows has great support, surprisingly, for TPM-backed sk keys using Windows Hello and OpenSSH. Protected with physical presence and anti-hammering at the hardware level, and easy to setup by just selecting a sk type key.

I only use password keys for things that need to be scripted.

You can just replace dropbear with openssh on OpenWRT. That was one of the first things I did, since DropBear also doesn't support hardware backed (sk) keys. Just move it to 2222 and disable the service.

I reenabled DB on that alt port when I did the recent major update, just in case, but it wasn't necessary. After the upgrade, OpenSSH was alive and ready.

GHCP also has magical rate limits that hit users that slam multi-agent workflows or other crazy request burners.

Mind you, I think GHCP is a great service at an excellent price, but the hardcore vibe coders complain about the rate limits that I've never personally experienced using the CLI.

Hey LLM, write me a system prompt that will avoid the common AI 'tells' or other idiosyncrasies that make it obvious that text or code output was generated by an AI/LLM. Use the referenced Wikipedia article as a must-avoid list, but do not consider it exhaustive. Add any derivations or modifications to these rules to catch 'likely' signals as well.

There, sorted!

The bigger question is: does Anthropic have a big enough moat to matter?

I've used/use both, and find them pretty comparable, as far as the actual model backing the tool. That wasn't the case 9 months ago, but the world changes quickly.

But onset of action is a very important distinction in medicine/pharmacology, as is dose.

Most abusers of methamphetamine are not taking it orally (slow route of administration) and are generally using much higher relative dosing than ADHD patients are using amphetamines. Potential for addiction and other physical harms are greatly affected by both of those things, so the comparison has some truth, but is obviously sensationalized.

Amor fati.

While stay at home parenting isn't, and shouldn't have to be, for everyone, it also isn't somehow a downgrade from being in the working world. If anything is doing something 'over and over', it's trudging to some job to push papers/keyboard keys around for 8+ hours.

This is the point missed by many. The trades are in high demand, right now, because of a labor shortage and demand from upper-middle class individuals without any DIY skills. A generation or two of pushing kids into college, and an almost disparaging view of 'getting your hands dirty' has built this perfect storm.

However, besides a few trades that use unions/licensure/apprenticeship as an artificial supply limit, most trades are only limited by a willingness to do the work. A few decades ago, trade work was much less expensive, because supply was higher and many did their own DIY, which limited what prices the market would tolerate.

Are they?

I'd put Codex 5.3 on par with CC for almost every task, and OAI has been rapidly updating their app, with a major initial release for Windows just a few weeks ago. Quotas are a moving target, but right now, Codex offers a better value by far, being very usable at the $20 level.

I don't have a dog in this race other than competition keeping them all honest. Claude led for so long, but I think that early lead has blinded many to how close it is now.

The only one really eating dust is Google. What a terrible offering. I wish it wasn't so, because they could really apply some price pressure to the competition with their scale and integration.

If you want to run EspHome inside HA, and you recompile the devices (every release of EH), you want a decent processor/disk. The ESP stuff is a surprisingly heavy compile for a puny microcontroller.

You're making it complicated with all the VLANs. HAOS in a VM (proxmox helper scripts for one-line install), and HA has plugins for all the other things.

Just deny WAN access to the IoT junk you don't trust at the router, or for things like cameras, a separate switch for those. That usually makes sense, since they're one of the few devices that must be powered with PoE and doesn't require gig+ bandwidth. A cheap 100mbit PoE switch will handle a good number of cameras.

I've always said one of the best non-major-related courses I took in college was Criminal Justice 101, which went through all the most applicable SCOTUS case law for common scenarios. Ignoring the variation in state laws, you could boil it down to about 30 rules of thumb. Many of the most important are covered in the classic YouTube lecture 'Don't talk to the police.'

Teaching this as a requires HS class would be an incredible benefit to society, because, on the flip side, many police encounters escalate to violence because the citizen has an incorrect understanding of where their rights end/don't exist.

The most obvious rule to follow is that you should always assert your rights (correct or incorrect) verbally only, as soon as you involve physical resistance, the situation will deteriorate rapidly (for you.) Any violations of your rights will be argued and dealt with in court, not on the street. Confirm requests/demands from officers are 'lawful orders', and then do them.

I'm curious why you don't like OpenWRT?

I've found it to be the better choice for x86 hardware, because it performs so much better on older CPUs. FreeBSD has gotten better with driver support, but the Linux kernel in OWRT is just a better base to build off of.

It worked for years for the SomethingAwful forums. A nominal charge for the ability to post, with plenty of 'timeout' chances for rehabilitation before an outright ban keeps out most of the junk.

It feels wrong at first to pay for commenting on a forum, but the alternative is almost always a gentle slide towards a trash dump. AI means that slide is almost a vertical slope.

I've found and fixed this bug before. There are 2 other ways to handle it

Dapper has a static configuration for things like TypeMappers, and you can change the default mapping for string to use varchar with: Dapper.SqlMapper.AddTypeMap(typeof(string),System.Data.DbType.AnsiString). I typically set that in the app startup, because I avoid NVARCHAR almost entirely (to save the extra byte per character, since I rarely need anything outside of ANSI.)

Or, one could use stored procedures. Assuming you take in a parameter that is the correct type for your indexed predicate, the conversion happens once when the SPROC is called, not done by the optimizer in the query.

I still have mixed feelings about overuse of SQL stored procedures, but this is a classic example of where on of their benefits is revealed: they are a defined interface for the database, where DB-specific types can be handled instead of polluting your code with specifics about your DB.

(This is also a problem for other type mismatches like DateTime/Date, numeric types, etc.)

This is exactly the argument that is (correctly) levied against firearm restrictions.

Even though I don't like the privacy implications, make sure you use the option to save and use past chats for context. After a few months of back and forth (hundreds of 'chat' sessions), the responses are much higher quality. It sometimes does 'callbacks' to things discussed in past chats, which are typically awkward non-sequiturs, but it does improve it overall.

When I play with it in 'temporary chat' mode that ignores past chats and personal context directives, the responses are the typical slop littered with emojis, worthless lists, and platitudes/sycophancy. It's as jarring as turning off your adblocker and seeing the garish ad trash everywhere.

From a 'public health' perspective, it makes perfect sense to limit the frequency of screening procedures by age and other broad risk-factors, but that doesn't help at the individual level if you fall on the unlucky side of those statistics.

Most cancers are still very much lethal once they progress to a certain point, and the best treatment we know of is early detection. Many of the cancer screens are harmless or don't add significant risk of death, so it really comes down to money and medical resource availability (also solved with money.)

I don't see much difference in someone paying out-of-pocket for a full-body MRI/colonoscopy vs. them spending way above average on any other item that slightly reduces the risk of dying (how many smoke alarms and fire extinguishers does your home have?)

Though ostensibly supportive of your claim, the first article says it best a few pages in (surprisingly honest):

"To date, there are no clearly established biological mechanisms that could explain the role of red and processed meat in the process of CRC carcinogenesis."

In other words, we see some small signal in epidemiological studies, and we want to speculate about mechanistic causes, even though this has been tried before to no success.

I think the 'elementary mistakes' in humans are far more common than confined to the mentally ill or intoxicated. There are entire shows/YT channels dedicated to grabbing a random person on the street and asking them a series of simple questions.

Often, these questions are pure-fact (who is the current US Vice President), but for some, the idea is that a young child can answer the questions better than an 'average' adult. These questions often play on the assumptions an adult might make that lead them astray, whereas a child/pre-teen answers the question correctly by having different assumptions or not assuming.

Presumably, even some of the worst (poorest performance) contestants in these shows (i.e. the ones selected for to provide humor for audiences) have jobs that require agency. I think it's more likely that most jobs/tasks either have extensive rules (and/or refer to rules defined elsewhere like in the legal system) or they have allowances for human error and ambiguity.

It's interesting how much focus there is on 'playing along' with any riddle or joke. This gives me some ideas for my personal context prompt to assure the LLM that I'm not trying to trick it or probe its ability to infer missing context.