Old Fucking Punk @lwriemen@librem.one

@lwriemen GNU/Linux is punk 😎🤟🏾

I know many software devs who use agentic coding at work only because they are forced to by their bosses, who are watching their token consumption. These devs hold the technology at arms’ length, clearly delineating a boundary for their own cognitive autonomy.

But don’t be fooled; that is not what most executives want them to do. Executives *want* their devs to surrender their cognition. They *want* devs to stop thinking about software problems, and let the agents take over, because they believe that’s where the elusive ROI is. They claim they want agents to “help” devs, but really, they want agents to *replace* them.

And they must. Because even the most conscientious software engineering executive must hang around the C-suite with non-software executives who are convinced that software engineering is obsolete. The sofware executive cannot push back without risking their career.

#ai

@ludicity I have to admit that vibe-coding *does* allow people to experiment with software faster (by non-consensually stealing labor from other people regardless of license or compensation, just so we’re clear). It’s hard to argue that evaporating a small lake to create a maybe-working clone of an app while you sleep can be pretty exhilirating, like stealing a TV from Best Buy and getting away with it.

The moral hazard here is that we conflate vibe-coding with software engineering, or more dangerously, that we conclude that understanding is *no longer necessary* to make software. And make no mistake, that latter point is often the “value proposition” that many executives see in AI: that it makes the entire field of software engineering obsolete, that they can simply get “product people” to generate software that they don’t need to understand.

The lure is strong. You start with small disposable scripts. Just ten lines of code, what’s the harm? But it never ends there.

#ai

There's an article going around[1] about how Anthropic's Glasswing project has ~38,000 findings (mostly still private) and only 442 CVEs.

While that's not particularly surprising to me (given my own experience with LLM vuln false positives) I feel like the analysis is lacking. We shouldn't lump together newer entries with older ones—instead, we should look only at the breakdown for entries that have had time to go through the full process!

And today bagder (unrelatedly?) pointed out something interesting—CVE assignment is badly backed up for at least one CNA. So... this might not be a good metric at all.

[1] https://www.vulncheck.com/blog/anthropic-glasswing-november-update

[2] https://mastodon.social/@bagder/117609225522066289

The more I observe Corporate America, the more dead-on accurate this article by @ludicity appears to be: https://hermit-tech.com/blog/ai-mania-is-eviscerating-global-decisionmaking

I watch demos of AI-enabled “tooling” and what have you, each of them promising much, but in reality asymptotically approaching what can only be described as “mediocrity, but automated” in practice. I hear people mention Claude or Copilot in conversations like they were name-dropping a celebrity to get the bouncer to let them in the club.

I see devs defer their opinion of pull requests to LLM summaries. I hear people talk about sending agent swarms to implement a feature that they barely understand. I see people give up on trying to understand architectures, preferring instead to sic swarms of subagents to implement a feature so they can close the ticket quickly.

Ironically, this is probably all reported as success in the executive dashboard. Look how agentic coding has improved engineering! How transformative it is!

And here I am, witnessing a mass lobotomy taking place.

#ai

@lwriemen @lrhodes well OS-Sci at least seems to disagree with you in this essay. CASA was suggested as an alternative view on software, one with a particular emphasis on community, and OS-Sci saw the need to say that such a view is incompatible with the four freedoms, especially with freedom zero.

Yup dependence on Microsoft Copilot *is* a major bummer: https://www.rnz.co.nz/news/business/2583930/microsoft-battling-major-outlook-outage #MSOffice363 #MSOutage

@lrhodes their essay demonstrated to me something i think i've seem mentioned before but never fully clicked for me: for many in the OSS movement the four freedoms are the end of moral reasoning. the CASA movement clashes with the four freedoms, therefore it's bad, QED.
this is especially apparent in this sentence from the conclusion:

> Freedom 0 is not a peripheral feature of FLOSS philosophy — it is the philosophy.

is there a need to examine the four freedoms? no, they're the four freedoms!

Paper is still punk.
Now thru 11/15, get 40% off select zines, pamphlets, and Comics here: https://www.pmpress.org/category/zine-sale

Seattle’s Massage Workers Demand Safety and Solidarity—Not Stigma: After police raids on more than 50 Asian massage parlors, workers are pushing for licensing that recognizes their expertise and an end to the “rescue” model.

Read more:
https://inthesetimes.com/article/massage-workers-labor-safety-rights-seattle-raids

@lrhodes It's always telling when "you should have morals" draws a critical response.

> But anyone who reads the manifesto alongside the classic texts of the Free Software Foundation and the Open Source Initiative quickly sees two world-views colliding. World-views that have formed the backbone of the internet for forty years.

Yes, it's pretty clear that the principles that have governed the internet for 40 years, and have seen it transform from a place of freedom and decentralized experimentation into one of repression, control and centralization *require* a challenge and a change.

The FSF, GNU and OSI are *spectacularly* failing to rise to the challenges of our age.

“The Work Ahead”

https://2ndbreakfast.audreywatters.com/the-work-ahead/

> It's clear – or I hope it's clearer – that a lot of voters actually hate "AI" – both the technology itself and everything that it symbolizes: wealth inequality, authoritarianism, corruption, nihilism, violence, theft, unaccountability, environmental destruction, indeed the destruction of any notion of the future other than the one that these terrible, terrible tech CEOs seem to have for us.

@alex said:

"In my twenties I also thought that forks were the worst, and why can't we just all get along and make the software. But now I know, it's not a zero sum game: If misogyny reigns, for example, there is just one broject. If we fork it, there are two projects. The broject is not harmed, because the misogynists can keep collaborating over there where as all decent people can collaborate on the second project. These people had left and now they can come back."

https://social.alexschroeder.ch/@alex/statuses/01M75HP65SV141A1N50D6DWTNG

Here is one of the worst ways you can use #AI for #software development: Have the machine build some stuff with zero #oversight or #accountability and then drop the responsibility onto the user.

This actively makes it worse and less secure for users. It's in the official @nextcloud app store.

#VibeCoding #genAI #AIslop #Knorke

Microsoft, one of the worlds most profitable companies, after laying off 23% of their staff in the past two years and offering 27% of their remaining staff packages to leave and then naming a product Microsoft Vibing, adding product features called Vibe, declaring Copilot writes a majority of code and going all in on vibe coding:

RE: https://mastodon.social/@mcc/117607410513383159

There is something profoundly telling in the fact that "I bet someone got a promotion for this" is one of the cruellest insults in this industry.