Slop, Slop, Slop

Slop, slop, slop. All I hear are people complaining that there’s too much “AI slop.” All that code everywhere on GitHub, online, bots everywhere. All that slop. I’m not sure anyone has really stopped to think about the amount of slop we already had.

How long has coding been around? Quite some time. We humans create code, and we often leave it “as is” and push out a prototype of something that’s usable but not the final thing. Software has gotten to the point where it’s never perfect and will always get updates. Something that helped with that was broadband internet. Back in the day, before broadband, there was what us old people call “dial-up,” and you’d dial into a BBS. It could take a long time just to get text to show up on the screen, let alone push any kind of software update. So you made sure your software did exactly what it was supposed to do, from the moment it was installed and the entire time it was used on the end user’s computer.

Broadband wasn’t just a home thing, either. We had it at work and at school. The enterprise had fast connections, but it was when broadband came to people’s homes that “software updates” started to enter the heads of programmers.

Mind you, we did have updates. For Windows, they were called “Service Packs,” and they usually came on a CD, or you did a fresh install of Windows. If there was an update you had to grab online, it had better be really important, because that download could take an hour. Then again, Windows can still take an hour to update today. As TechRepublic reported, Microsoft released an update in April with fixes for 165 vulnerabilities, including what seem to be two zero-days. Not good.

Some people, including Steve Gibson from Security Now, believe Anthropic’s Mythos had a hand in that. Anthropic gave around 50 companies, including Microsoft, Apple, and the Linux Foundation, access to Mythos, and Gibson thinks that’s why we saw such a big Windows update. He also doesn’t think Anthropic is bluffing when it says the model is too powerful to release to the public yet.

Mythos also found a 27-year-old bug in OpenBSD, one of the most secure operating systems out there and one that a lot of firewalls run on. The point I’m trying to make is that humans aren’t perfect at making code. Don’t get me wrong, humans are very good at making code. Look how far we’ve gotten today, and what we’ve built with what we’ve learned. But now we’ve made something that can learn, in seconds, what we programmed these computers to do, and these things can produce really good stuff. ONLY if there are humans overseeing the code and the project.

What’s also changing is the process of making the things we need. We no longer need to know how to write Python or Bash scripts. We just need to know how to be a good project manager. What you’ll go to school for now is how to manage a team. If you can manage a team, could you imagine? Think about it. You have all these ideas for a company or a project. So what do you do? You build a team. AI. It’s now AI.

“AI” is possibly the wrong term for it, since it’s such a broad term. AI can be spell check, or language translation. What you’re really using is an AI assistant that talks to agents. Each “agent” is like a person:

Agent 1: You do this.
Agent 2: You do that.

And so on. But keep in mind that each agent is using an LLM, and that counts as usage depending on the system you’re using. If you’re using a provider like GitHub or Claude, your usage just goes up. If you’re hosting it yourself with LM Studio or AnythingLLM, the requests just queue up, and each one waits for the one before it to finish.

I think we’re in a transition to something really big, and we’re right in the middle of it. We’re learning how we program and make software all over again. Since the dawn of coding, it’s been the same: code was everywhere and we really didn’t have any standards. Over time, we created those standards and started using them. Now we’re learning the new standards for AI coding. Because honestly, computers are better at talking to themselves and just making the machine do whatever it needs to get the job done.

Thank God we don’t all have one of those home robots that takes 45 minutes to do anything, or we’d be up a creek. Well, I don’t think it actually takes 45 minutes. I think I just heard it takes a while. Anyways...

The companies that relied on software as their main source of revenue are in a lot of trouble, because now we can all pretty much make our own software. Software that might not be perfect for others, but may be perfect for you, and may only need to be updated when... well, never.