You can certainly have your opinion. But calling people that don't agree with you 'fucking trash' shows a lot more about you than it does them.

"Upon which, I will re-change it to whatever I want, anyway"

Are regressing back to superstition and pseudo-science. Bigfoot doesn't exist.

AI Overviews was already kinda worthless if you were looking up anything that had to do with human behavior, but I noticed reddit becoming a primary source, and every interaction quickly became like this:

"Why is this X thing happening?"

(Some useless statistic that shows people to be little more than brain-damaged monoliths with no uniqueness whatsoever)

"That can't be right, that statistic has to be wrong"

"You are right to call that out. What is actually happening is (Additional useless statistic that simply shows the other side of an issue being the actual thing happening)

"You just said that X was how things are done, why are you suddenly going the other direction?"

"There are varied and hotly debated topics in this space..." ( thenproceeds to spew useless statistic based on reddit nonsense)

Whatever.

Probably somewhere along the ditch where real RAG with properly managed self-hosted models have blown past them. Ollama + gui is not much of a workflow. You need some type of embedding/vector db system, a web search engine/scraper so the model can make tool calls to the web, some type of hybrid search with a re-ranking model, and then the final generation model. A typical stack is going to be something like openweb-ui + litellm + vllm on the backend.

Or yeah, just use Claude Desktop and Anthropics stack, because the big providers are doing this for you already to some degree. Not trying to piss on the parade, but firing up ollama with a small Qwen model or whatever is not a comparison.

...like they do with Opus?

Opus is a nice help when trying to get past a coding problem, but during high-demand periods, the output of Opus declines so much that it becomes unusable. It reasons in circles, and starts outputting code that is one step above nonsense, and then can't live-update artifacts anymore, so you blow through you session in minutes, when it should take hours.

Just make it once a day, or hey, every hour. That way, I can update Chrome between the gig of Win11 updates coming in.

I love how someone expressing their opinions is slowly moderated down to Troll. Heaven forbid that someone don't agree with you.

In my experience, it just ends up requiring replacement of the starter battery more often, and the batteries required are larger, AGM type batteries that tend to be more expensive than smaller/flooded lead acid type batteries, (if you care about battery cost).

Now, I suppose, if you sit in traffic a lot, and don't need A/C or heat too much, don't have a lot of accessories that are on a lot, (like heated or cooled seats, radio, etc.), and are absolutely wanting to save that last $2/month on gas, then go ahead. All I would ask is that the button *remember* the setting, between on/off ignition cycles, so if I turn it off, it just stays Off. I have installed a few harness adapters that connect to the harness going to the start/stop button so that it can 'remember' it's setting and only enables if the button is pressed to enable it back on. That is how it should have been from the factory.

Agree. That could be solved by the AI-bros actually charging for the services, which will have to happen someday. They will have to start making actual profit on this shit, someday. Although, I'd also think that a lot of the useless shit people ask AI to do could be handled by local models, which would be cheaper to run. I would think summarizing an email should certainly not require 10 top of the line GPUs to do by now.

No problem for the AI hyper scalers...they'll just do some 100 trillion dollar deals back and forth and it all will work out.

For a Beemer, it's the Pretentious Dickbag Subscription. For another $799.99/month, we'll put a middle finger where the turn signals would normally be.

If I was going to give a public interface to anything, it's going to go through the same process that I have been using for a number of years now - a templated design that has been tested/pentested/tested again for holes. And these days, it will likely be behind an AWS api gateway as well if we're talking about internet-facing apis. It won't be some EC2 instance with a public ip and a sec group of 80 -> 0.0.0.0/0.

I should have added in my original post that I feel AI tools can be very useful in the hands of trained and experienced developers. The CEO that vibe-codes some application over a weekend and then wants it pushed to prod at his company on Monday is not what I am talking about here. We have a group now at my work that has vibe-coded a very complex, very expensive to run app with an insanely complex DB structure that will be unmaintainable. I have already made my concerns known about that particular project, and it will not go to prod until the people that want to assume liability for it sign off. It is an MCP implementation that has security issues all through it. Yes, that shit isn't going to prod until it has been eval'd by competent engineers and/or someone wants to take the fall for it.

So, what I am saying, is that I agree with you. AI has it's place.

I have spent the last 2 weeks really getting into CoPilot for helping write code, (python specifically). I am very fluent in Python, and have been for awhile. I give a shit about architecture, clean/simple code, data types, and can pass my stuff through mypy and it passes.

I started two greenfield projects - one for embedded development and one for math modeling using numpy and python, (C was a long time ago for me, and I am not deeply familiar with numpy, but I have knowledge of calculus and advanced algebra and how it works on paper). My knowledge of embedded dev is very basic.

Within a week, I had working code for both that was much faster than I could have done by hand. Now, the code *works*, but I need to dive deeper into how *clean* the code is. But I also started with a basic architecture diagram of what I wanted, and that is also a big factor. And I think that that is the most important thing - the architecture. I don't believe it makes sense to just start blasting prompts at an agent and then just blindly run whatever it gives you. I start with a foundation, I start developing the pieces of the architecture, and I read through the AI responses and push back on anything that I don't want, (and when it adds interesting things that I find useful, I work that into the architecture). But I work in small blocks.

AI coding does make a difference. But it isn't replacing *architecture* anytime soon. If you really want to spend hours and hours and hours building syntax and having to run the program 3000 times a day while you debug where a semicolon is missing, then great. I used to say that I'd never leave infrastructure, too. And now that I work as a cloud architect/dev, I am happy to never have to make another run to the datacenter 100 miles away at 2 a.m.

It is what is is. AI does have some good uses. We just need to figure out how to not destroy the environment completely, fill the internet with misinformation, or block new engineers from entering the field, and/or put Sam Altman or Elon Musk in charge of the world in the process.

Google is your friend, bud.