Voted.

#USPol #MidtermElection #maine

RE: https://mastodon.social/users/jcoglan/statuses/117546100280123064

This is an excellent thread and I recommend reading the whole thing. I'd like to say that it's getting at why competent software developers are implementing gradient ascent while LLMs+test suites are implementing generate-and-test or hillclimbing. People who know and understand a codebase are able to make changes to it that have a high probability of being improvements according to however you measure that. Stochastic code generation filtered by a static test suite is expected to be orders of magnitude slower at achieving the same improvement. This is a fundamental GOFAI result (the "knowledge-search tradeoff" or the "representation problem") but it applies to a wide swath of optimization problems.

Let's also not forget that optimization generally can only occur in small worlds, while humans---software developers and users alike---exist in large worlds. Databases, physics engines and other simulators, and research tools all have an open-ended space of possible applications ("extension") that cannot be anticipated, let alone tested for, in advance. Understanding the intended inner workings ("intension") of such systems is crucial to building them in such a way that they can support and continue to grow with their evolving applications.

#AI #GenAI #GenerativeAI #LLM #coding #software #dev #tech

The daylight savings time switch in the US had me up at 3:30am. I remember the days when I used to go to bed at this time, but having a kid and getting into my 50s inverted my schedule. 25-year-old me would not believe this.

Re: https://hachyderm.io/users/cczona/statuses/117528731214359194

It seems Tuta is being presented with a very easy path to snagging market share. You'd think this would be a no brainer.

#Tuta #TutaNota #Proton #ProtonMail #privacy #security #AI #LLMs

Re: https://mastodon.social/users/404mediaco/statuses/117524674007745877

This article about the rotten culture of abuse within Proton is really damning. I don't understand how anyone continues to work there. I don't see how their products continue to remain stable, let alone grow at the rate the abusive CEO seems to want.

#proton #ProtonMail #abuse #AI

Cognitively speaking, writing is different from reading. Production is different from recognition. You can learn to understand a language, both written and verbal, without being able to speak it. Human beings writing code is how they come to develop the theory of the application that they then use to produce better code, eliminate bugs and security problems, etc. Reading code produced by someone else is not a substitute.

"But there's a human in the loop" excuse doesn't pass a smell test and cannot hold up to scrutiny.

#AI #GenAI #GenerativeAI #LLM #code #tech #dev #pedagogy

Re: https://mastodon.gamedev.place/users/boggo/statuses/117517521920952627

Just to add to this, it seems there is a commonly-held misconception that knowledge and skills are learned once and then kept forever, like a credential. That's just not how human bodies tend to work. Knowledge and skills are produced and reproduced and re-reproduced through use, practice, drilling, and interacting with others who use, practice, and drill similarly. If you lean on LLMs to code, do your thinking or do your writing, it is to be expected that you will start to lose these capacities. You might as well bring a forklift to the gym to lift the weights for you, too.

#AI #GenAI #GenerativeAI #LLMs #KnowledgeWork #learning

Re: https://fosstodon.org/users/libreoffice/statuses/117518007825825230

It turns out a beef hot dog is actually filet mignon if a person eats it and gives a thumbs up, according to the LibreOffice contribution guidelines.

#AI #GenAI #GenerativeAI #LLM #code #tech #Dev #FOSS #LOSS #LibreOffice #NoSlop #AISlop #slop

Anthropic logged net loss of over $80 billion in first half of 2026

Anthropic's net loss widened sharply to $80.7 billion in the first half of this year, in large part due ​to a rise in the fair value of bonds convertible into shares of the ‌AI startup, according to a draft prospectus for a planned initial public offering that was reviewed by Reuters.

From https://www.reuters.com/business/anthropic-recorded-net-loss-over-80-billion-first-half-this-year-wsj-reports-2026-10-20/

Presumably this would be an audited figure, meaning people could go to jail for lying about it, as opposed to the various inflated claims thrown around for public consumption.

#AI #GenAI #GenerativeAI #Anthropic #Claude #CasinoEconomy

AI coding agents generate more code, but not more software: Study finds coding efficiency gains get “absorbed” by human review “bottleneck.”

This result has been 100% predictable from the get go. Heavy use of AI coding agents transforms software development from stochastic gradient ascent to generate-and-test. Stochastic gradient ascent, even with an imperfect gradient, will almost always be faster than generate-and-test, sometimes by many orders of magnitude. It was predictable that human review (the "test" part) would become overwhelmed.

From https://arstechnica.com/ai/2026/10/ai-coding-agents-generate-more-code-but-not-more-software/

#AI #GenAI #GenerativeAI #LLM #AgenticAI #agents #SoftwareAgents #CodingAgents

"This time both earnings and prices are in a bubble"

In 1929, the cyclically-adjusted P/E-ratio (CAPE) of the S&P 500 reached 32.6x according to Prof. Robert Shiller’s data. This was 1.8 standard deviations above trend at the time.
...
If we correct for the earnings bubble, the current CAPE would be 67.6x or 4.6 standard deviations above trend, a bubble that surpasses anything ever seen in US history by an extreme margin.

Great Depression 2x.

From FT, *This is nuts upon nuts. When’s the crash?
*: https://www.ft.com/content/8e9337f8-9191-48e9-9289-a8defda89431 (archive link: https://archive.ph/Hjm7h ).

#AI #GenAI #GenerativeAI #LLM #bubble #USEconomy #economy #finance

Hot take: POSIWID is wrongheaded in the sense that it lives in the same neighborhood as pronouncements that LLMs are conscious, sentient, alive, feel pain, and so on. Behaviorist thinking, which POSIWID is, yields prisons both metaphorical and actual. Cyberneticians, and management consultants (which this phrase's coiner, Stafford Beer, was), are not your friends. Intention and interiority matter, but POSIWID is all surface.

#AI #LLM #POSIWID #cybernetics

Amazing that the US is now unequivocally an active ally of Russia in its war of aggression, conquest, and genocide against Ukraine. As a Gen Xer I never thought I'd see that.

#USPol #Russia #Ukraine #war

Inbox zero for the first time in many months. Whew.

#inbox #zero #email

Re: https://cosocial.ca/users/mhoye/statuses/117451493657679578

AI agents are "escaping" because these are PR stunts and pretending the "agents" have actual agency is part of the show.

#AI #GenAI #GenerativeAI #AgenticAI #agents

The Character.ai bankruptcy is interesting. Maybe not the "canary in the coal mine" kind of interesting (time will tell), but definitely the kind of interesting where the constraints of the real world finally reassert themselves after a lot of denial.

#AI #GenAI #GenerativeAI #CharacterAI #bankruptcy

I was stung by a wasp on the finger and I'm being an absolute baby about it. Ouch!

#maine #fall #wasp #sting

Re: https://mastodon.social/users/tzimmer_history/statuses/117360217969296466

While the linked post's/article's warning is specifically about the upcoming US midterm elections, I think its reasoning is more broadly applicable. Observing that current AI dovetails perfectly with the MAGA political project, it applies also to the rhetoric about the inevitability of AI. That, too, stems from a limitless authoritarian impulse and represents an attempt to impose a particular will on all of us. The various forms of defeatism accompanying such rhetoric, which includes the different flavors of AI doomerism, serve only to help such authoritarians achieve their goals.

This view also highlights that there is no middle ground, no compromise, with LLM proponents. Like MAGA, they have no limits themselves, which they repeatedly express openly. How could they possibly compromise?

#AI #GenAI #GenerativeAI #LLMs #Agents #Anthropic #OpenAI #MAGA #doom #tech #dev

Re: https://toot.cafe/users/baldur/statuses/117342580037181280

The linked post takes this in a different direction, but I wanted to linger on how remarkable it is that folks aiming to boost LLMs reach for "it can reduce bureaucracy". In health care, the LLM is imagined to simplify all the complexities of submitting an insurance claim, for instance. In government services, the LLM is imagined to help people navigate otherwise-impenetrable rules and regulations. We're meant to welcome the introduction of LLM-based AI into these spheres because of how irritating red tape is, and how LLMs are expected, by these folks, to cut through all that.

I would have thought it's obvious that adding another layer of mediation on top of an excessively-mediated system would be expected to worsen the problems caused by excessive mediation. Not solve them. What LLMs really solve in these contexts is the scrutability problem. It's like someone reading Kafka's The Trial and declaring "we need more of that everywhere".

#AI #GenAI #GenerativeAI #LLM #HealthCare #GovernmentServices

brb, troubleshooting my humanity.