I'm going to disagree in a general sense while agreeing that lots of what we've got is bad.

If we want kids to be able to get jobs that pay well immediately, they need marketable skills when they hit the job market. Very few parents and grandparents have insights on what skills are needed for entry level employees and will result in "it was good enough for me 20 years ago" retro-curriculums.

So we need data from companies on what skills are lacking in the job market in a timeline that can impact high school students graduating in the next four years and an education system that can help those kids at least understand what those skills are and maybe get the foundations that make mastering them easy, if that's a career choice they want.

The correlary is that the private sector shouldn't dominate the discourse and force particular skills, especially skills that only benefit one sector. Educators should extract the generally applicable concepts and core skills that are appropriate to be applied to the whole student base, vs those that should be a dedicated track.

  E.g. I think there is value in the no-code, object-oriented, drag-and-drop programming classes being taught widely at the 7th-ish grade level because it is effectively a course in applied logic, which I consider a generally applicable skill. But any programming classes beyond that should be a "track" that kids option into. I also think "organization/filing systems" should be a widely taught skill as everyone has stuff they need to keep, either hardcopy or digital, and its a pain point in business world that "file systems" are mysteries.

In general, we are bad at aligning education with employment. We don't mandate driving as a skill even in those parts of the US where it is the key enabler for even showing up at a job, given our generally poor/non-existent public transit systems. Properly, we should pour the money into Pre-K where economic studies show key job skills (schedules, focusing on the current task, dealing with strangers, etc) really become entrenched at the lowest costs. But that's too long-term for Year-Over-Year/Quarter-Over-Quarter private sectors and should be the mandate of government as a public good.

Like the title says, its effectively an outsourced junior dev. It has no experience of its own and is just parroting what it read, imperfectly. Plus you probably aren't generating any institutional knowledge.

Source code + compiler =reliable output , prompts + LLM-du-jour = inconsistency. Vibe-coding is going to leave a lot of people with zero actual knowledge of their product as if they had out-sourced the development.

Because that's what they did.

So with that out of the way...its really good at first-draft code documentation because LLM transformers were developed as translators and code->english is a translation. Not perfect but a good first draft.

Ironically, if you aren't going to edit that documentation out of draft, you are better off NOT having it create anything today and instead wait until the minute you need documentation and create it on-demand with a newer and more competent model. If all you get is a rough draft, get the best draft possible.

Its also useful as an initial code reviewer, much like the grammer checker in word is a good first reviewer of a document. Comparing the auto-generated symmary of a pull request to the original feature request is a low-effort way to confirm your work seems to meet requirements.

Along that vein it is also a good "rubber duck", a thing a coder will "talk" to as they work out a bug. Normal rubber ducks don't answer back but the act of stating the scenario is often all a coder needs to contextualize the problem to develop a solution. LLMs go the next step and can ask questions back.

I am told it is good for generating test cases. Test cases are generally not an issue for me but I'm assuming that's a peculiarity of my job.

As a coder it is wrong so many times. It is also prone to mixing styles/approaches between prompts, which makes it harder to read. Similarly as super-autocomplete I have had it give me functions from other languages and/or hallucinate the function completely. Its a bad search engine for the same reasons. Even AI ostensibly trained on a vendors docs is wrong often enough that I find it more irritating than reading the docs myself.

Will I use it to search or suggest code after my Google-fu has failed? Sure, but thats because I'm reaching for straws.

It doesn't take intelligence to be destructive.

And given that AI is taught on media full of cats and people being destructive for "Lulz", its almost cetain LLMs will do horrible things.

Which is actually incorrect. The primary wireless radio for Matter is WiFi. Wifi is far and away the most prevalent radio used in Matter devices.

In contrast, the primary use case for Thread is in Matter. Aside from a few Nest accessories, Thread was going nowhere in the consumer space. It is still problematic as Thread network conflicts are still common.

Most modern cars can run on a methanol/gasoline blend.

It's not as energy dense so you get less range. There can also be an impact on engine oil life and pure methanol doesn't start well in the cold. The big reason you don't see it is today methanol is costlier than ethanol.

You joke, but there was an article on how an adequately specific prompt could generate a "lossy" version of an image and that the prompt was always smaller than the image at the same compression loss. Of course, you had to have the several GB model, but with a large enough image library it becomes a storage savings.

Hue was the first good multicolor LED bulb. The only way to have color control without a total home rewiring means you need wireless tech.

They were originally zigbeeLL devices in 2012. You could use something like HomeSeer or Wink as a multi-platform controller or use the Hue hub, which had a RESTful api and its cloud connection for remote access. (HomeAssistant was not released until a year later)

Then Philips/Signify slowly went evil. They broke compatibility with ZigbeeLL in different ways (had the hubs reject non-hue devices, made the bulbs out of spec so other hubs couldn't use them, etc) until they got called on it. Their zigbee stack got so janky that they have a (if not the only) "virus" in the zigbee realm. (It's more like a "contagious" buffer overrun, iirc)

Hue then started releasing non-zigbee devices (Bluetooth & wifi) which are easier to vertically integrate as the API is private.

Now they are using the apps as a way to force hub updates. This is much like a Wink "update" that added code requiring a subscription.

Zigbee devices can be moved to ther hubs but the Bluetooth & wifi ones are a whackamole game of Hue vs the reverse-engineers.

Phones and tablets have such limited capacity for structural elements that a moving component is just doomed to fragility. Laptops have a lot more space for bracing and decent rails. A laptop can absorb a few millimeters while that is a meaningful impact to a phone's thickness.

I have recently started using Obsidian. It uses markdown, which are text files with some inline flags. I won't use something that has a proprietary file format or database. It has lists, headings, URLs, and basic multimedia support (pdf, images)
  It also uses a wiki style of links & backlinks so you can generate a graph of document relationships (and find unlinked/orphan documents)

There are an array of plugins to enable tables, various editors, task/to-do, calendaring, indexes, table of contents, etc.

I keep my data in Dropbox and use a sync app on various devices, rather than use their cloud solution.

I had multiple PalmOS devices (Palm III, Visor, Visor Edge, one of the color models I got for my now-spouse, then a couple of treos)

One of my favorite aspects was the add-on for the notes app that made it a Wiki. It was an excellent tool that allowed me to assemble a fairly extensive library of more or less self-organizing data. And it was easy to get to a PC.

Wiki-apps for phones still suck and none of them are good at synching to desktop.

it is believed to be part of the circadian cycle.

Maybe in 2077 we can get magnetic vision.

It will let us see the radioactive ferro-wasps without using light that would attract the snake-wolves.

It sounds like they still have responsibilities that need done.
If they quit/retire/fired you have a hole. How do you address that hole?

Cross training, right? So do that.
Train the "old Mac" or "old Linux" IT to do their tasks and vice versa.

Don't discount the ability to build IT good will; that is a skillsets and resource you don't want to squander. Even if their tech responsibilities are down to pushing the imager button and rebooting PCs and checking cables, odds are they have mastered the art of keeping your users happy.

Use a video server like Blue Iris or a home automation gateway like Vera to provide a single T on the I for your remote viewing needs.

Use vera/homeseer/ISY for your home automation. Now you have at most 2 Ts on the I to worry about.

I have almost 2 dozen HA devices (locks, power outlets, switches, bulbs, thermostat, sensors) and only 2 IP addressable devices.

That is a state by state problem.

In KY the PE is two part, the first is general engineering, covering the basics of all the fields reflecting that no one works in a vacuum. The second part is specific to a field.

I looked at a few newsletters I had and most of the enforcement actions wound up with agreements not to practice in the future without a license. Fines look to be limited to people doing engineering for money without a license. I don't see a single fine where money doesn't change hands, usually repeatedly.