My desire to use Mastodon is basically at zero now because post-"Claude", hanging out here means inevitably I'm going to have a conversation with someone who tolerates, or even uses, "generative AI". And what's the point of being in a community where that's a risk. Interestingly* there is absolutely no chance of this on Bluesky, because there are artists there

* And oddly, given how "AI"-brained the *admins* there are

No longer interested in talking about open source unless the conversation starts with how to build an alternate open source community without "AI code assistant" users contributing to it
@mcc Sadly the Copilot monster came to #FreeBSD on GitHub last week.
@bms48 @mcc Wait, how did copilot "come" to FreeBSD ?
@trashheap @mcc The LLM-driven entity is submitting AI slop bug reports and "reviews".

@bms48 @mcc Ewww...

Last I heard FreeBSD was working on a draft policy. that was leaning towards actually banning LLM code. (This was discussed at BSDCan last year). That hasn't changed has it?

@trashheap @bms48 last I heard netbsd had done this but don't cite me
@mcc @trashheap @bms48 last summer FreeBSD leaned strongly towards no (see https://reviews.freebsd.org/D50650 ), but the more recent "accepted" change is weaker (see https://reviews.freebsd.org/D54817 ). I don't like the weasel words in the newer proposal, and the proposed "developer's guide" doc (which you have to be logged in to see) is overlong and makes it too easy to miss the actual point.
@fedward @mcc @trashheap You're not going to like this, but there are FreeBSD folk who are looking to leverage Generative AI, but in a very limited and targeted way, e.g. automated code reviews. We would aim to do it in a controlled way away from Microsoft's monopolistic interest as Douglas #Rushkoff predicted would happen. It isn't a strict go/no-go area. The UK CDPA act does offer copyright protection to GenAI content but the key test is originality (system prompts must be very specific).
@bms48 @fedward @trashheap My goal is to have no "genAI"-emitted code on my computer. If FreeBSD uses the lie machine to review code but not to write it, then I'd consider using FreeBSD, but would possibly not send PRs back upstream (the environmental impact of LLMs can be heavy and I'd prefer not to take any action that could result in an LLM query) and the idea of an open source project I can't contribute back up to is weird.
@bms48 @fedward @trashheap This said tho there's the bigger problem of long term trust. Once "AI" is inside an organization it spreads. If you've accepted this is a real technology which is okay to use and the question is just where and how much is appropriate, then "how much" will increase until it's on the table everywhere for everything. The people in the org who already want to use more of it will win more and more arguments over time because hey, you agreed to it *last* time, why not now.
@bms48 @fedward @trashheap I've already switched from Windows to Linux to get away from "AI" features and now have to switch away from Linux for the same reason. Each switch is disruptive. If FreeBSD is just going to be like Linux and just be two years *slower* in adopting "AI" than the other thing, it's not clear why I bother with FreeBSD. It's not like I'm interested in your OS on the merits. If I cared about product merits and not the ethical/political dimension I'd probs still be using Mac.
@mcc @fedward @trashheap I hate to keep invoking Robert Heinlein here, but he said it best: TANSTAAFL: There ain't no such thing as a free lunch. And Atkinson’s Casket, Sept. 1833, has ‘The price of liberty is eternal vigilance.’
@bms48 @fedward @trashheap Okay, well possibly the price of liberty is not using FreeBSD. And maybe if I go with NetBSD instead of FreeBSD I don't waste two years learning something I'm going to throw away.
@mcc @fedward @trashheap To be frank, everything I know/do and action items wind up in a genai-todo.txt file I haven't found time to action yet. I am tracking the Cloudflare C++ developers, macOS guys at Apple too, who seem to have success with Claude under Opencode.ai which is a bit of a sandbox. The thought of stuff like Codex, Cursor or OpenClaw just scares the hell out of me, it's too much INFOSEC and OPSEC risk to use those tools. Cursor forks VS Code and that is just a big ball of crap.
@mcc @fedward @trashheap From Ed Zitron's latest post today: https://www.wheresyoured.at/the-ai-industry-is-lying-to-you/ "I’m sure there are software engineers using these models ethically, who read all the code, who have complete industry over it and use it as a means of handling very specific units of work that they have complete industry over." This. This is what we are aiming at, and how the Cloudflare team wield it, and how I intend to wield it. But Huang wants you to take out a mortgage and keep vibe-coding.
The AI Industry Is Lying To You

Hi! If you like this piece and want to support my independent reporting and analysis, why not subscribe to my premium newsletter? It’s $70 a year, or $7 a month, and in return you get a weekly newsletter that’s usually anywhere from 5000 to 18,000 words, including

Ed Zitron's Where's Your Ed At
@mcc @fedward @trashheap So there is a real argument to be made here that LLMs are the technology of mean reversion. They favour what they have been trained on. They are stochastic parrots in the sense that @emilymbender has been warning us all about. When such a thing is wielded in a very targeted way, it might yield a small economic benefit, but again, this largely accrues to the elite. I have code in iOS and macOS myself, so I probably do qualify as that; I am not your average bear.