Honey bees ARE an invasive species in North America. Invasive species are bad, right? /s

Paper mills are truly awful to be around. Necessary, but awful.

I think you really hit the nail on the head. The most successes I have had with LLM code is when it's building on an established foundation, a well-structued database, well-structued MVC set up, or whatever. If you're using a well-documented framework, that's another plus.

If you know enough to guide the AI in the direction you want it to go, you're far more likely to get good results. Heck, I've had good luck with just writing a function prototype and having it fill in the guts. I've had good luck telling it to refactor so-and-so class according to whatever principles. I use it regularly with great success to write elaborate SQL from plain english prompts.

It feels very much like the end of an era. I do not by any means think software development as a job is going away--I'm not even sure jobs will ultimately be lost--but it will never be like it was before, for better or for worse.

I only started using Claude Code a few months ago, and you are absolutely correct about the cli / code-integrated tools.

I had Claude translate a COBOL (esque -- DATABUS) program into a modern language and framework today. The plan phase took about 6 minutes, I made a few edits to the plan, and writing portion took about 4 minutes. I got claude to run some tests comparing outputs, and they were identical. I then myself ran similar tests and got the same results. Pretty neat.

I hate having to tweak the legacy DATABUS code and recompile. The subroutine / goto method of programming is hard to wrap my head around now.

To be fair, Kasparov came back and won in 1999 and earned a 6 game draw in 2003. There was a period where the grandmasters were figuring out weaknesses in the AI play.

I can't say I've followed chess at all recently, but my impression is that humans are no longer competitive.

Now do Go.

30 years ago Go was considered almost an almost impossible problem for an AI program to compete at even a high amateur level.
20 years ago Go programs started being able to beat strong amateurs / weak professionals
10 years ago AlphaGo decisively beat the best Go players

We're in a situation where improvements in the performance of AI system are linked to both more advanced techniques and massive increases in compute power. I don't see either one stopping any time soon.

Progress can be scary.

Is that all AI's fault? I also don't know how bad the job market for beginning coders is!

I graduated from undergrad a bit more than 20 years ago with a computer science degree. At the time there were less than 100 majors per year. This was roughly 4-6% of the student body. Comp sci was well behind economics, public policy, biology, political science, and maybe some others in terms of popularity.

Starting in the 2010s, the number of computer science majors started to grow very rapidly. In 2024 there were almost 500 majors! Almost 1/3 of the entire student body was a computer science major.

This is a national university that historically was not especially a "tech" school. I thought the quality of computer science students in the early 2000s was not especially high. I was the only person in my introductory 80+ people required class who had any experience in asm. I was able to take a couple of grad level classes, and my experience there was that most of the students were H1Bs (or related) who were simply there for more credentializiation. Many undergrad students talked about they just wanted to get their degree and then move straight into management. Yuck. (This is all one of the reasons my primary career today is not as a developer!)

In any case, it seems absolutely insane to me that my school is spitting out that many majors. I'll be curious to see the 2025 and 2026 numbers...

It's really not the same thing. See my other post https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Flinux.slashdot.org%2Fcomments.pl%3Fsid%3D23961458%26amp;cid=66085502

I don't know how it will work with all the different firewalling options, etc., with Linux, but that's what makes it special for macOS (which comes with lsof, tcpdump, etc.)

It's really nice software. You don't just get a visualization of current connections, you can get popup of new notifications AND the option to set up incoming/outgoing rules. Something like "Firefox is attempting to access slashdot.org port 443:

Allow Once
Allow Firefox to connect to slashdot.org port 443 any time
Allow Firefox to connect to any server port 443 any time
Allow Firefox to connect to any server, any port any time
Deny Firefox all connections

When I got my first Mac laptop, around 2004, Little Snitch was one of the first pieces of software I bought. One of the nice things about the Mac at the time (and historically through OS 7 / 8 / 9 days) was there was a very strong small developer writing shareware culture. That's disappeared a bit more recently (in large part due to the rise of open source, Linux ecosystems, and so forth), but Little Snitch is still great. Another example, Adium remains one of the best chat clients I've ever used. BBEdit is very solid text editor that goes back to 1992! Etc.

That's the part that perplexes me. I can't say I have followed the Vision Pro that closely recently, but at least in the early days, it was very locked down. There are large parts of the system (sensors) that I understand are still not accessible to 3rd party apps. I remember reading several VR developers who were trying to port software from the Quest to AVP and simply couldn't.

To me it would have made sense to open it up entirely. Porn, video games, emulators, whatever!

Are you unsure what "agentic" means? Generically, agency means, more or less, being able to do things. An agentic AI program (ChatGPT in agentic mode, OpenClaw, Claude code) can take actions without being controlled by humans. This is also sometimes called an autonomous agent, but "agentic" has become the dominant term over last year or two.

If you were to try something like Claude code, for example, you would see that it can run shell commands, grep through a source tree, edit files, run git commands, compile, execute, review output, edit code, compile again, etc.

Your post is confused as it assumes that LLMs looking for security vulnerabilities are -- in your words -- "LLMs can unreliably find some defective code patterns if they are obvious enough. (Remember, they cannot do deduction, just statistical pattern recognition. Too much noise or too far from the template and they fail."

That is false. The LLM models in an agentic system are doing more than just looking for "defective code patterns."

LLMs can perform source code level analysis, but they can also, in agentic mode, run fuzzing tools, scan for open ports, write a custom program to attempt to fuzz or exploit an open port, review the output, modify the code to try again, compile, repeat, and so forth. Multiple instances of agentic AIs can do this for many hours and more.

As I've said multiple times before Gweihir, I really don't know if you're trolling or not. You've seemed at least somewhat informed before, so I'm really surprised you didn't know what agentic means, or what has been possible with these tools. ChatGPT agent mode has been out for about a year, and Claude code for a bit longer.

VisionOS 26.4 added support for foveated streaming. It wouldn't surprise me if Apple and Steam have been working together on this.

Then again, Apple's treatment of the Vision Pro perplexes me, so who knows what Apple is doing there...

Nice to know that Greg Kroah-Hartman doesn't know what's talking about--you should let him know that you've cracked that nut, since he is actively working on understanding the situation.

That is not an accurate statement. The current round of discussions are centered around projects that in large part are agentic. Source code analysis is only one of the detection methods that is being used.

Disregarding the furthest extreme radicals on any side of a philosophical debate may generally be a good thing! Though, to be fair, sometimes it is the radical extremists that drive things forawrd. Kind of like the Overton Window..

Look, it's not complicated. Disregard everything that Sam Altman says. Disregard both the furthest extremes of the "AGI is here and sentient" / "LLMs are Godlike!" and the "LLMs are trash that are not useful and aren't going to improve and are a passing fad" (gweihir). All of the above are not insightful.

Everything has changed. ChatGPT-3 was released in 2022. Everything HAS changed since then, and LLM technology and models have improved dramatically in the last 4 years. Why would you not expect statements like "that used to be true, but not anymore" to be problematic? Humans couldn't fly, until they we could. We couldn't visit the moon, until we could. Changing your baseline assumptions is part of human progress.