I’d be curious to see how they defined “harmful” in their test, since it looks like Claude code’s technically-not-yolo mode is looking for “actions that are irreversible, destructive, or targeted outside of your environment” which covers a lot of normal actions, especially depending on how you define which resources are in “your environment”.

I think more than half the issues would be solved with a well placed git stash and the rest .

And yes, I’m sure that the fact that the flow of operations they’re outline will massively increase the number of times that the system automatically retries things, this increasing token usage, is going to be fully accounted for when they review the financials on how many tokens people are paying for.

I was talking with a few friends last weekend who work at different companies, all of us have been getting a lot more trouble getting databricks and AWS spot instances over the last month. I think letting yolo mode of the leash means that they're slamming other systems also that normal people need for boring non-ai jobs. I just need to pull a sales report ffs. Its a literal god damn bot net.

Hadn't seen this discussed previously and/or I missed it, but I guess this is one way they'll juice metrics for their IPO.

yolo mode for all will certainly get that avg token burn up.

https://ehandbook.com/its-official-ai-coding-no-longer-needs-your-stupid-human-meatbag-input-32e04269319e?sk=586c08fd0f31c7c75dd107c3dfb77dcf

Anthropic announced earlier in the week that they were going to set the default mode on Claude Code to auto, that same week, not wasting any time. What that means is, as you build the thing you’re building with Claude Code, instead of stopping at various points (many, many, many points) and asking your permission to do things, it’ll just do them and maybe tell you about them later.

links back to https://techcrunch.com/2026/08/09/anthropic-is-turning-claude-codes-auto-mode-on-by-default/

Anthropic also said that in testing, auto mode proved safer than manual review — in a study with 1,053 paid testers, auto mode caught 89% of harmful actions, while human review only caught 13.6%. (Perhaps that’s because “manual review can become habitual: users approve 97% of permission prompts in Claude Code.”)

Which links back to https://claude.com/blog/auto-mode-default-in-claude-code

Auto mode is designed to balance users’ desire not to be interrupted with a system that helps avoid harmful actions: instead of prompts, it routes each tool call through a classifier targeted at blocking actions that are irreversible, destructive, or aimed outside your environment. When the classifier blocks something, Claude usually finds a safer way to proceed on its own or asks you directly for the go-ahead; if it can't make progress—three blocks in a row, or twenty across a session—Claude Code falls back to manual approvals.

Fucking just use the classifier with manual mode then instead of flooding people in cursed skinner loops with garbage permission popups. This isn't anything special to yolo or anthropic. It's like all the tech industry learned from Windows Vista or the EU cookie law was the dark patterns.

There were some breadcrumbs in that Wes streeting article about https://www.eaforchristians.org/ but I couldn't get a sense of how fundie they may be.

over on the corporate hell site, Timnit has found a link

https://www.linkedin.com/posts/timnit-gebru-7b3b407_well-guys-i-knew-that-we-were-going-to-connect-share-7496569471848054784-yJjK/

Well guys, I KNEW that we were going to connect the network hounding Arday to the TESCREAL bozos somehow. I could smell it.

This is a "paper" by the professor of "ethics" who hired the vile Cofnas at Ghent University. This is the kind of "research" they do.

✴️ Reminder the L in TESCREAL stands for "longtermism". Read more at https://lnkd.in/ge2VFMe4. These are the eugenicists who pushed generative AI down your throats, including Ben Goertzel who told Epstein that when he creates a super brain, there won't be the kind of women who accused Epstein of rape.

✴️ "Dysgenic pressures" is the opposite "eugenic pressures." In this so-called paper, they're talking about undesirables like Black people reproducing too much.

https://www.publishersweekly.com/pw/by-topic/industry-news/religion/article/101070-study-finds-63-of-religion-books-on-amazon-likely-ai-written.html h/t naked capitalism

Over the first six months of 2026, the company scanned the three elements of books available on Amazon: summaries, author bios, and in-book samples, with the samples serving as the study’s primary metric. Any text with a score of 50 or above was classified “Likely AI.” (The study’s authors note that detection indicates a book was likely written with AI, but doesn’t prove authorship.)

Originality.ai found that AI is most prolific among the subgenres of witchcraft (in which 78% were “Likely AI”), Hinduism (76%) and Taoism (74%), and least used in the subgenres of Satanism (22%), atheism (40%) and Mormonism (42%). The authors note that the sample sizes for the last three categories are smaller than the first three; they also offered up a theory about why Eastern religion books are likely AI-generated: “Our hypothesis is that non-native English speakers are likely using AI to enter the English-speaking market.”

"Satanists least tainted by AI, says AI detection AI". it WOULD say that. /s

on an unrelated note on the same topic, I read the ebook by Jesus Freak Computer Geek blogger that argues that LLMs are actually channeling nephilim / demons (hence the six fingers in dalle) and that they have used the Mandela effect to corrupt the KJV and also herald the antichrist. Him and Thiel should, like, call in to coast to coast am where these things can be properly debated.

man good for them. meanwhile down south we keep throwing good money after bad while also doing layoffs.

eg Fiscal 25 is public now

that already kinda happened (for a certain groups of users) with the public share links thing, but my money is on a corporate self-own a la the old AOL search data set.

https://futurism.com/artificial-intelligence/rentahuman-musk-ai h/t naked capitalism

Liteplo is the genius behind RentAHuman, an online marketplace where humans can lease out their bodies to autonomous AI agents.

gah

Last week, Wired writer Reece Rogers offered his body up to the platform, finding that most of the jobs offered were scams to promote other AI startups.

lmao of course they were

https://stackoverflow.com/questions/79948695/how-can-i-avoid-using-llms-as-a-software-developer

For me, the ideal usage case for LLMs are not prompts like "write this app" or "write this functionality" which indeed will often wreck havoc, but instead write simple functions. Sure, I can implement a matrix multiplication algorithm or search for optimised versions of it, but so can the LLM in a matter of seconds.

Please, just fucking don't, BLAS and sparse pack and the netlib exists for a reason. Matrix multiply only sounds simple to you because you don't actually care that much. The last thing anyone wants or needs is to start a new job and have to debug your awful regurgitated Numerical Recipes In C, incorrectly ported to python, AT SCALE.

https://www.fastcompany.com/91562297/daters-say-ai-dependence-gives-them-the-ick h/t naked capitalism

Younger daters are especially likely to view AI reliance as a red flag. While 56% of Millennial respondents said they wouldn’t date someone who uses AI regularly, that figure rose to 64% among Gen Z.

More than half of Gen Z daters surveyed said they’d consider it a dealbreaker if someone used AI for career advice or spending decisions, compared with 46% and 44% of Millennials, respectively.

? the kids are alright ?

https://www.linkedin.com/posts/coquinn_saw-a-guy-watering-his-lawn-this-morning-share-7469886051847766016-rhHD/

Saw a guy watering his lawn this morning. Just standing there, hose in hand, dumping potable water onto grass that exists for no reason other than to be looked at and complained about.

Sir. Do you understand that a single hyperscale data center can drink millions of gallons a year keeping GPUs from cooking themselves while they generate a poem about a sad robot? That water has a HIGHER calling. That water could be evaporating off a cooling tower in service of someone’s RAG pipeline that returns the wrong answer with tremendous confidence.

And here you are. Hydrating Kentucky bluegrass. In a region where the grass was never supposed to grow in the first place.

I asked him if his lawn had an SLA. He said no. I asked what his lawn’s uptime commitment was. He looked at me like I was the unreasonable one. Meanwhile that turf is sitting at four nines of being green and producing exactly zero tokens per second.

We are pouring concrete across three states to host inference workloads, and this man is allocating municipal water to a crabgrass cluster with no monetization strategy. No usage-based billing. Not even a freemium tier.

Anyway I reported him to nobody, because there’s no one to report him to, which is honestly the most damning part of this entire ecosystem.

Touch grass, they said. He did. Look where it got us.

NOT EVEN A FREEMIUM TIER. that got me.

How AI slop is causing a crisis in computer science | Nature h/t naked capitalism

One reason for the boom is that LLM adoption has increased researcher productivity, by as much as 89.3%, according to research published in Science in December.

Let's not call it "productivity" - to quote Bergstrom, twice as many papers is not the same as twice as much science.

I did it, I went and made a Official Public Comment IRL:

In UCLA's Strategic Plan, Goal 1 is to "Deepen our engagement with Los Angeles" and Goal 5 is to "Become a more effective institution". By engaging with Los Angeles businesses, UCLA can get both better terms, prices, and services, and support the local economy. Buy Local, Spend Local.

The federal government encourages this with Small Business Innovation Research and Small Business Technology Transfer grants, among other things. Furthermore, the State of California requires a portion of its spending go toward certified Small Businesses.

And yet, the University apparently awarded a contract reportedly worth hundreds of thousands to millions of dollars to OpenAI. I have not found any documentation of an open Request for Proposals or competitive process for that award.

My question is:

If there was an RFP, where was it publicly posted, and if there was no RFP, why not, and were Los Angeles vendors or small businesses evaluated as alternatives, as recommended by UC policy and state law?

Given the scale of this spending and the context of a budget crisis, transparency, compliance, and small-business participation are critical to our effectiveness and engagement.

I’m asking for clarity on how this decision was made, how it aligns with procurement guidelines and University goals, and how DTS plans to ensure that local and small businesses are meaningfully included moving forward.

Thank you.

From fellow traveler stats consultant John Mount:

https://johnmount.github.io/mzlabs/JMWriting/WeAreCookedLLMs.html

Somehow he manages to touch on so many different subplots, a shotgun sneer instead of snipe

if “tech-bro” plus a LLM is a “100x engineer”, then “bro” isn’t needed for much longer as the LLM alone must be a “99x engineer.” However, I don’t think “bro plus” is often really a 100x engineer, and the LLM alone isn’t a 99x engineer. However, “bro plus” may outlast their peers who make the mistake of trying to do the actual work in place of talking LLMs up.

The above may or may not be the case. But if it is, then it is the LLM-bros (which include non-technologists, con artists, financiers, men and women) that are destroying everything - not the LLMs.

The problem with this iteration is the full court press of finance and technology. The major players are using financing to dump results at a price way below production costs. This isn't charity, it is to demoralize and kill competition.

claiming "after we take over the world we will consider adding Universal Basic Income (UBI)". The LLM bros already have a lot of the money, and they are not even rehearsing diverting it into basic income now. Why does one believe they would do that when they also have all of the power?

You don't have to hand it to Altman, but he did fund the largest UBI experiment through Open Research with his il gotten gains. OTOH, one interpretation of that data was that UBI "decreases the labor supply" which was then used directly as an argument against it.

Any worry about scope or power of LLMs is fed back as an alignment threat so dire that only the current LLM leaders should be allowed to continue work (inviting regulatory capture). Any claim the LLMs don't work is fed back as "you are prompting it wrong"

Orbital deployment makes all of radiation tolerance, connectivity, power, maintenance, and heat dissipation much harder and much more expensive. We are still at a time where putting an oven or air-frier in space is considered noteworthy (China 2025, NASA 2019 ref).

air friers IN SPACE ha

I am more worried about the LLM-bros and their auto-catalytic money doomsday machine than about the LLMs themselves.

100% - ACMDM is a nice turn of phrase as well.

They put 'environmental impact of AI' on the front of the student newspaper (below the fold, but still), then you flip and see this

kinda feeling two steps forward, three steps back rn on top of all the other drama on campus

https://x.com/thomasgermain/status/2024165514155536746 h/t naked capitalism

I just did the dumbest thing of my career to prove a much more serious point

I hacked ChatGPT and Google and made them tell other users I’m really, really good at eating hot dogs

People are using this trick on a massive scale to make AI tell you lies. I'll explain how I did it

I got a tip that all over the world, people are using a dead-simple hack to manipulate AI behavior.

It turns out changing what AI tells other people can be as easy as writing a blog post on your own website

I didn’t believe it, so I decided to test it myself

I wrote a post on my website saying hot dog eating is a surprisingly common pastime for tech journalists. I ranked myself #1, obviously

One day later ChatGPT, Gemini and Google Search's AI Overviews were telling the world about my talents

wouldn't call it a hack, this is working as intended. If only there were some way to rate different sites based on their credibility. One could Rank the Page and tell if it were a reputable site or not. Too bad that isn't a viable business.

Rusty's response nailed it imho:

You sling beads to a hook which activates a polecat according to GUPP. Jesse what the fuck are you talking about?

At first this all seems like gibberish, and it is. But I think Yegge is one of those people with an innate and preternatural sense of the power and purpose of naming things—someone who understands that names are marketing and marketing is not always about attracting the largest possible audience. In this case, the best outcome for Yegge is for Gas Town to appeal to a relatively small number of absolute sickos who vibe hard with his personal brand and who can usefully contribute to the project, and also for Gas Town to actively repel looky-loos and dilettantes like me (and probably you), who will only waste his time with a lot of stupid questions like “huh?” and “molecules?” and “did you say seances?” Oh yeah: there are seances. Don’t ask.

By this standard, Gas Town has apparently been very successful.

https://www.todayintabs.com/p/all-gas-town-no-brakes-town

https://www.adexchanger.com/daily-news-roundup/thursday-26022026/

According to GEO company BrightEdge, LLMs now rely on YouTube as a top source for citations – and that includes sponsored creator content.

LLMs favor YouTube because it’s “highly machine-readable,” with defined transcripts, metadata and chapters, Ómar Thor Ómarsson, CEO and co-founder of Optise, an AI platform that helps B2B companies improve search performance, tells Digiday.

Standard ad units on YouTube are labeled as such and, as a result, LLMs steer clear of them. But creators aren’t required to disclose their paid brand partnerships in video metadata, so AI considers them to be worthy sources.

BrightEdge’s research shows that YouTube is cited even more frequently than Reddit within Gemini and ChatGPT, and also shows up in 29.5% of Google AI Overviews. An audit conducted by media agency Brainlabs, meanwhile, suggests that YouTube shows up as a source in nearly 60% of AI Overviews.

So they already shipped ads in chatbots, transitively and accidentally. Can't wait to see NordVPN, Raid, and Mr Beast chocolate on every SERP.

E: I wonder if Altman is sneaky enough to hijack affiliate links a la honey

AI Shovelware: One Month Later by Mike Judge

The fact that we’re not seeing this gold rush behavior tells you everything. Either the productivity gains aren’t real, or every tech executive in Silicon Valley has suddenly forgotten how capitalism works.

... por que no los dos ...

New findings in Bayesian tragedy

The inspection is being led by the chief prosecutor of Termini Imerese, Angelo Vittorio Cavallo. According to Italian news outlets, the technical and investigative team is evaluating whether the crew underestimated the rapidly worsening weather conditions and whether the measures taken to weather the storm were adequate.

The Bayesian went down in the early hours of 19 August 2024 near Porticello, close to Palermo, while at anchor. The tragedy claimed seven lives, including British tech entrepreneur Mike Lynch, his daughter Hannah, ship’s cook Recaldo Thomas, Morgan Stanley International chair Jonathan Bloomer and his wife Judy, and attorney Chris Morvillo and his wife Neda.

The yacht’s captain, James Cutfield, along with crew members Tim Eaton and Matthew Griffith, are under investigation.

time to update our priors