This is exactly what CLAUDE.md, GEMINI.md and AGENTS.md (or copilot-instructions.md) are for. You put your requirements, instructions, guardrails and notes in there. My general flow for things I just want to rip out is to put my core requirements into Gemini Deep Research and ask it to flesh them out (the code assist "plan" modes do the same thing but Deep Research is usually a little better), give it a good once or twice over to see what it got wrong, add guardrails based off of previous experience with the coding agent (e.g. do not use this library, this class of functions, this approach, etc. etc.) and then drop it in the context file. The resulting code is going to be substantially better than if you just give it a simple prompt. If you find something its continually screwing up, add another guardrail or not to the context file.

If you're not doing this, you are doing it very, very wrong and you're going to get garbage. You'll still get garbage with a context file but a lot less of it and generally it's not completely off the rails.

They have the vowels but are still working on the full translation.

"remote control" as in turn off the bus... not "drive the bus".

The statement from the Yutong could be a little weasel worded. The article is talking about remote deactivation, the spokesperson is talking about data-collection. Nothing in the quoted statement addresses remote control. Chinese companies have a history of doing this when responding to this type of thing. 'A' is broken. What are you talking about, 'B' is just fine... nothing to see here! They misdirect or just flat out lie (Anker with their Robovacs being a recent, good example).

All Slot I/II's required heatsinks and most had fans (some OEM's didn't but it was intended for the OEM to install the fan). Now the heatsink was often preinstalled (or part of) the cartridge... maybe that's what you were thinking? The max TDP was around 20-30W, not crazy but still required a fan or a chonky passive heatsink. The card/slot was also not done for cooling reasons, it was done so they could bundle L2 cache with a dedicated bus instead of having it on the motherboard (L2 still wasn't on the chip package at this point).

Your statement about how LLMs work in general is 1000% correct and critical for people to understand. If you have a decent understanding of how they work, you start to understand the importance of good prompts and guardrails which have a significant impact on the quality of the output.

However, "it does NOT make any reasonable attempt to verify" is no longer true. Reasoning models do make some efforts to verify, in some cases it's pretty significant. Gemini Deep Research outputs it's "reasoning" and you will frequently see it go down a rabbit hole only to come back and say "nope, that's wrong". That doesn't mean it's always right and doesn't make mistakes but I pretty much only use Deep Research/Think now and save Pro for the simple things. Gemini Pro is definitely worth the money. I'm still on the fence about Deep Think, haven't had it really blow my hair back but maybe I'm not throwing the right class of problem(s) at it.

It's a sad, sad world we live in.

You do realize that other countries have tariffs as well? And yes, some of the are retaliatory but most of them were in place long before Trump was in office.

https://www.customs.go.jp/engl...
https://www.customs.go.jp/engl... (from 2020)
https://ec.europa.eu/taxation_...

derp. Fine, fine, I'll own that mistake.

At least in Japan. The "deals" are "you save 40% on this item we marked up 39.999999%", deliveries are often delayed by days with no notification or reason and especially lately is more "you'll get it when you get it.". It's gotten to where if I know I need something, I just go to the store and buy it.

A weird side effect of this is that family names are dying off in Japan. While I doubt it will just be Sato as this professor is predicting, there is a really short list of common last names and other ones stand out (thankfully the different kanji for the names helps a bit).

Obligatory: How many a**holes we got on this ship anyhow? Yo! I knew it, I'm surrounded by a**holes.

I know you're joking but Jupiter is actually the fastest spin planet (in our solar system). It completes a rotation in less than 10 hours, which is crazy fast given it's size.

I work with plenty of people who are vegan for religious and cultural reasons. They will bring it up at opportune moments, we're discussing where to go out and they'll ask if there are any vegan options at the proposed restaurants.

Then there are those who "discovered" veganism and form who being vegan is the religion... they will let you know it any chance they get and will generally question your dietary choices as to boot. I believe the poster was referring to the latter.

Not sure if you're serious or trolling but testing is a separate step in the build process, you'll have a build target in the Makefile and a test target (typically broken into multiple sub steps). With something as big as the linux kernel each component/module is going to have it's own build and test steps that have dependencies on each other and all fall under the top level build and test. Let's say there are 100 components that have a dependency on hdrtest. If you want to test your component(s), their tests are going to fire off when you do your build. That's going to slow down your build and testing of your component (and piss you off... grr...).

It's not that the linux kernel doesn't test, they just organize and execute the tests in a way that makes sense for a project of that scale.

There are two groups that are facilitating this:
1. The c-suite types who are driven by profit and just see the numbers not the scope of the data. While your data is worth something, it's not a lot.
2. The engineers doing it who abstract it as a challenge/problem to be solved rather than something impacting people. And frankly don't always realize what they are doing. I've had more than a few cases were well intentioned engineers brought me a solution that was revealing far more than they intended.
3. Okay... #3... the sociopathic a-holes in both groups who genuinely don't care

Abstracting it out... the goal is to provide better data about you. Good data is worth more. Your friends and family provide more data points about you. If all my friends are buying GTA VI, chances are I'm going to as well ... so start serving me GTA adds. A family member is having a baby (I swear Amazon knew my wife was pregnant before we did)... start popping up ads about baby products.

That's where the dissonance comes it. It's viewed as a technical problem (how do we serve the best ads) instead of a social one. Probably the most effective way to stop it will be to require employees of companies that collect data to submit the complete/maximum data for themselves. That snaps it back into reality. There will still be those in group #3 trying to weasel word their way out of it but hopefully enough staff will start thinking about how they can do it better.

For what it's worth, I worked for a fintech that you'd see countless threads online discussing how we were obviously collecting and selling all the data. We actually weren't and every engineer I worked with was trying their best to protect privacy. #2 did happen from time to time but teams got better about trying to identify it before it happened and would fix it if was brought to their attention. Not everyone or every company is bad.