Like all automations it's not 100% perfect, and they all require a human to monitor them.

Turning over a problem to a LLM as solved means we won't be able to do the task anymore. Because we will never learn how. We will not know that the LLM has made a mistake, and will be unable to fix it ourselves.

Right now we can correct the LLM, as the years go on we will become unable to do so.

If you look at certain documentations, we lost already how we d othings.

USA forgot how to produce things.

This will not prevent us from doing it.

Most people who drive a car have no idea how an engine works. I know and agree it's a scary future, but I wonder if it's really _essential_ to understand things on a deeper level.

Most people are quite capable of understanding higher-order ideas without fully grasping the underlying details too.

It is essential that _someone_ knows how that car works.

Maybe less than we might hope. Not only is the knowledge distributed over the hundreds of people involved in designing a car, there may be conventions/habits of design and manufacture whose purpose has become hazy but are nevertheless important, or details of a design iterated upon whose original motivation is less understood.

I find at the heart of all this AI slop is a deep sadness at the loss of connection. It used to be when I read something, or saw something that I knew it was real. Someone created it for me to experience. It was a form of connection.

But now I'm no longer certain. Unless I really know the source, have met them in person, or seen them talk I just don't know if what I'm trying to connect with is real.

Even then when using AI just for myself I limit it to building tools, looking for errors. But when it comes to full blown generation, I just don't like what it makes.

I feel like an outlier in all of this. But isn't this just more AI slop? How is this different from text generation or image generation?

Like many people I have used AI to generate crap I really don't care about. I need an image. Generate something like, whatever. Great hey a good looking image! No that's done I can do something I find more interesting to do.

But it's slop. The image does not fit the context. Its just off. And you can tell that no one really cared.

This isn't good.

The difference is that coding agents can run the code that they produce, fix any bugs, build tests and generally demonstrate that it works.

You can't do that for images and text.

I'm a skeptic, because Self Driving is sold as a digital chauffeur.

Not 99% of a chauffeur, 100%. (or 99.99999%)

The roll out of this is clearly limited by the number of remote employees that are filling in the 1%.

For me AI has given that back to me. I'm back to just getting stuff built, not getting stuck for long when working in a new area. And best of all using AI for cleanup! Generate some tests, refactor common code. The boring corporate stuff.

I'm using Claude code to help me learn Godot game programming.

One interesting thing is that Claude will not tell me if I'm following the wrong path. It will just make the requested change to the best of its ability.

For example a Tower Defence game I'm making I wanted to keep turret position state in an AStarGrid2D. It produced code to do this, but became harder and harder to follow as I went on. It's only after watching more tutorials I figured out I was asking for the wrong thing. (TileMapLayer is a much better choice)

LLMs still suffer from Garbage in Garbage out.

don't use LLMs for Godot game programming.

edit: Major engine changes have occurred after the models were trained, so you will often be given code that refers to nonexistent constants and functions and which is not aware of useful new features.

before coding I just ask the model "what are the best practices in this industry to solve this problem? what tools/libraries/approaches people use?

after coding I ask it "review the code, do you see any for which there are common libraries implementing it? are there ways to make it more idiomatic?"

you can also ask it "this is an idea on how to solve it that somebody told me, what do you think about it, are there better ways?"

> before coding I just ask the model "what are the best practices in this industry to solve this problem? what tools/libraries/approaches people use?

Just for the fun of it, and so you lose your "virginity" so to speak, next time when the magic machine gives you the answer about "what it thinks", tell it its wrong in a strict language and scold it for misleading you. Tell it to give you the "real" best practices instead of what it spat out. Then sit back and marvel at the machine saying you were right and that it had mislead you. Producing a completely, somewhat, or slightly different answer (you never know what you get on the slot machine).

Both the before and after are better done manually. What you are describing is fine for the heck of it (I‘ve vibe coded a whisper related rust port today without having any actual rust skills), but I’d never use fully vibed software in production. That’s irresponsible in multiple ways.

Do you also light candles and chant?

I wonder how the police state works into this. China spends a lot of money on internal surveillance and suppression of information.

If the belief was genuine, would this be necessary?

> big software projects I've seen succeed were down to a few inspired leaders and/or engineers who set the tone. People with emotional intelligence, tact, clear vision, ability to really gather requirements and work with the end users. Leaders who treated their staff with dignity and respect.

I completely agree. I would just like to add that this only works where the inspired leaders are properly incentivized!

After learning that the Amazon Go store was power by hundreds of people watching video because the AI could not handle it was a real eye opener for me.

Is this why Waymo is slow to expand, not enough remote drivers?

Maybe that is where we need to be focused, better remote driving?

Waymo does not believe that remote drivers are responsive enough to be able to safely operate. Safety drivers communicate with the self-driving system, and can set waypoints etc. for the navigation system, but the delays inherent make it unsafe, is what the Waymo people say publicly at least.

The reason that Waymo is slow to expand is that they have to carefully and extensively LiDAR map every single road of their operating area before they can open up service in an area. Then while operating they simply do a difference algo on what each LiDAR sees at the moment and the truth data they have stored, and boom, anything that can potentially move pops right out. It works, it just takes a lot of prep- and a lot of people to keep on top of things too. For example, while my kid's school was doing construction they refused to drop off in the parking lot, but when the construction ended they became willing. So there must be a human who is monitoring construction zones across the metro area, and marking up on their internal maps when areas are off limits.

> Maybe that is where we need to be focused, better remote driving?

I think maybe we can and should focus on both. Better remote driving can be extended into other equipment operations as well - remote control of excavators and other construction equipment. Imagine road construction, or building projects, being able to be done remotely while we wait for better automation to develop.

This is an interesting idea. What are the expected benefits? Off the top of my head:

* Saves on commute or travel time.

* Job sites no longer need to provide housing for workers.

* Allows the vehicles to stay in operation continuously, currently they shut down for breaks.

* With automation multiple vehicles could be operated at once.

The biggest benefits seem to be in resource extraction but I believe the vehicles there are already highly automated. At least the haul trucks.