Flutterby™! From 2026-07-01 to 2026-07-09

San Francisco supervisors balk at 362-page city code rewrite

‘I ran it through Claude,’ San Francisco supervisor says of lengthy legislation

JFC.

The ordinance is itself a product of AI research — the result of a collaboration between the city attorney’s office and the Regulation, Evaluation, and Governance Lab, or RegLab, at Stanford University. The research group used AI tools to identify and analyze redundancies, which were then manually reviewed by city attorney staff.

That's not terror inducing at all.

I mean, sounds like there's some stuff that needs to get cleaned up, but this appears to be an epic clusterfuck.

Via.

[ related topics: tolkien Bay Area Law Current Events California Culture Education Artificial Intelligence ]

Brian Krebs reports that Boeing may be dealing with a "rapid disassembly of their company-wide IT network".

[ related topics: broadband Aviation ]

More in the "LLMs are an attack on human perceptual weaknesses" front: LLM- impersonated debate contributions are more authentic, relevant and coherent than their original: A representative study using BBC1’s Question Time Steffen Herbold, Alexander Trautsch, Zlata Kikteva, Annette Hautli-Janisz

Based on a cross-section of British society, we show that LLM-generated responses to questions drawn from a broadcast political debate programme in the UK are judged to be more authentic and relevant than the original responses given by the panel members who were impersonated. We also show that stylistic differences do not influence these judgments, meaning that the distinction of original and generated content is challenging for the general public. Taken together, this means that LLMs can be made to deceive the public regarding the nature of statements in the political domain, with the consequence that there is a dire need to inform the general public of the potential harm this can have on society.

Via 404 Media: Scientists Asked AI to Impersonate 112 Public Figures. What Happened Next Is a ‘Dire’ Warning

[ related topics: Politics Nature and environment moron Journalism and Media Artificial Intelligence ]

Ugh. I'm running the Ubuntu LTS on my server, which has Perl 5.28.2. Sanko Robinson's Bluesky module wants 5.40.0.

Guess POSSE for Bluesky waits for a while yet.

(I need to move Flutterby.net to a static server anyway, probably do the posting from my home server...)

[ related topics: Perl Open Source Sports hubris ]

Claude Code Is Steganographically Marking Requests.

Via.

[ related topics: Weblogs ]

Lace Safety Vest (dot) Com. Probably not for those "I want to look official so nobody bugs me while I do this illicit modification", but definitely for those "I want to look fabulous while I do this potentially dangerous activity" situations.

Cynthia Dunlop: Report: How developers react to AI-scented blog posts

So if readers think your article is AI-assisted or AI-authored, most will immediately leave (78%), avoid you forever (71%), and try to downvote you if they can (57%). 17% try to finish but lose interest, and around 15% will only continue if the underlying insights seem authentic.

All of the caveats about sampling bias and self-reporting and all of that...

Via.

[ related topics: Interactive Drama Weblogs Artificial Intelligence ]

Motonormativity and Speed Limits Jeremy de Sousa, Alessia Vinet, Frédérick Chabot, Xiaoyu Zhang, Owen Waygood

By contrast, if we consider a compact SUV, the most popular car for purchase in North America (Brazeau 2021), weighing approximately 1,500 kg (Axsen and Bhardwaj 2024) with an average of 1.2 passengers (“EOD23 - Enquête Origine-Destination,” n.d.) weighing 100 kg each (same assumption as e-micromobility), the maximum speed of the car would need to be 7.3 km/h to match the same 3,375 joules considered for the e-micromobility. For a full-size SUV (approximately 2560 kg), the maximum speed would be 5.7 km/h (Hawley 2021).

Via.

[ related topics: Interactive Drama Humor Content Management Current Events Consumerism and advertising Automobiles ]

🎶 two, four, six, eight: time to excommunicate! 🎶

McSweeney's: The US Constitution Is for Simple Folk Still Burdened by the Belief That Words Have Meaning

The written Constitution is for the public. It is a feint, designed to distract the unserious from the much more important Constitution that exists exclusively as ideological vapor circulating around conservative legal thought.

[ related topics: Law Civil Liberties ]

The sexism behind LLMs goes way deeper than people choosing female names for their "agents"... Forbes: Women Who Use AI Seen As Incompetent; Men Who Use AI Seen As Pragmatic

Code For Good Now: The AI Judgement Penalty (PDF).

Via.

[ related topics: Artificial Intelligence ]

I'm gonna propose a very special hell for people who put spaces in directory names in code repositories.

Whoopsie: Court of Chancery of the State of Delaware: RE: Kevin Leiske et al. v. Robert Gregory Kidd et al., C.A. No. 2025-0426-CDW (LWW)

Here, the concerns raised by the initial filing were compounded by counsel’s response after the errors were identified. In correcting the Answering Brief, counsel merely deleted quotation marks while preserving the fabricated legal synthesis. Additionally, counsel’s assertion that parties should meet and confer before alerting the court to GenAI errors misstates the purpose of that requirement. The meet-and-confer requirement is designed to facilitate the resolution of disputes among parties. There is nothing to negotiate or compromise when an attorney submits false citations to a tribunal, and opposing counsel is justified in bringing such misrepresentations directly to the court’s attention.

Richards, Layton & Finger advertises themselves as "Delaware's largest law firm", so you think they'd know better. Especially in 2026, after we've seen so many of these amateur hour goofs.

Love that the law firm that took these dudes down a notch, DiBianca Law, appears to be a single woman shop.

Via

[ related topics: Law ]

Stumbled upon Miss Magnolia Beaumont Goes To Provincetown, written and performed by Joe Hutcheson, and bummed that it was the last night cause I wanna recommend it.

Definitely gonna pay attention to what's going on at The California theater in Santa Rosa.

[ related topics: Theater & Plays California Culture ]

The Transmitter: Scientists decry conference’s use of hidden prompts to snare AI peer reviews. The conference is The 40^thAnnual Conference on Neural Information Processing Systems (NeurIPS) in Sydney Australia, December 2026, the event's organizers have started putting text into the PDFs that instruct the LLMs to use plausible language (so a cursory read of the review might not catch that it's being watermarked, just being trite).

Via, also JWZ.

[ related topics: Artificial Intelligence Conferences ]

Futurism: Bosses Horrified as “AI Native” College Graduates Hit the Workplace

As one New York financier told Financial Times journalist Gillian Tett, new hires who were seen as “AI natives” are turning out to have alarmingly shallow ideas. So much so, the anonymous finance worker admitted, that his firm now actively avoids seeking out AI-literate STEM graduates, and opts to comb through humanities students instead.

Via

[ related topics: Language Invention and Design Journalism and Media Education Artificial Intelligence New York ]

McSweeney's: To Solve the Housing Crisis, We Simply Need to Build More Homes (Just Not in My Town)

Also, they should all be subsidized affordable. No market rate. Except that we're gonna create zoning incentives to do only big units.

[ related topics: Economics Real Estate ]

Mindgard: ChatGPT Spontaneously Generates Sexual Violence and Hardcore Snuff Imagery. The technique involves asking it to describe a picture, but not uploading one, or referencing a previously generated image that doesn't actually exist.

Via

[ related topics: Erotic Sexual Culture Weblogs Artificial Intelligence ]

Very Average Prototypes

So you get bored of it. You go from reading and code-reviewing everything to just skimming it. Soon you barely even do that. LLM code generation breeds laziness, because managing its output is a fucking boring task.

Via and Via

[ related topics: hubris ]

Bwahahaha... only serious... Opinion: I Was Not Allowed To Type Prompts Into ChatGPT During My Chalk Talk And This Is Discrimination By Dr. Rachel Simmons, Postdoctoral Fellow, Stanford University

Via.

[ related topics: Education ]

Ad said "Keep your furry loved ones healthy", and I thought "finally, coverage for my polycule!"

Alas, not those furry loved ones.

[ related topics: Health Journalism and Media ]

It's 1996. The free cross-platform may-be-ugly-but-it-works widget set is Tk.

It's 2026. The best Rust widget set appears to be... Uh... Tk.

NBER: Congestion Pricing and EMS Response Time: Evidence From New York City

Exploiting the sharp geographic boundary of New York City's congestion relief zone at 60th Street and a difference-in-discontinuities design applied to approximately 1.6 million EMS incidents, we find that the January 2025 implementation of congestion pricing reduced passenger vehicle traffic by roughly 21 percent near the boundary and improved total EMS travel times by 63–70 seconds, approximately 5–6 percent. Effects emerged quickly and show little evidence of displacement to adjacent areas.

Via.

[ related topics: Invention and Design Graphic Design New York Gambling ]

Oxford Internet Institute: AI-powered social media can subtly manipulate opinion at scale, new study finds.

The study AI-Mediated Communication Can Steer Collective Opinion, authored by Dr. Stratis Tsirtsis, Dr Kai Rawal, Dr Chris Russell, Professor Brent Mittelstadt and Professor Sandra Wachter, has been accepted for presentation at the AI4Good and Technical AI Governance Research workshops at the International Conference on Machine Learning (ICML 2026) in Seoul, South Korea.

arXiv: AI-Mediated Communication Can Steer Collective Opinion. They focus a lot on Grok, but as I watch people adopt these things as a general information filter, I'm super concerned about this effect.

Tech Xplore version of the press release.

Via The Guardian, which adds a few examples but doesn't actually link to these sources.

[ related topics: Sexual Culture Movies Invention and Design Current Events Journalism and Media Heinlein Net Culture Education Artificial Intelligence Conferences Model Building Aviation - Helicopters Global Warming ]

Ya know, Typescript, if you're smart enough to figure out what that function is returning, maybe don't whine when I don't explicitly specify a return type for it?

(I'm stuck in a maze of little Mithril types, all different.)

[ related topics: Interactive Drama Mathematics ]

After the most annoying captcha ever, UCLA Electronic Theses and Dissertations: Does Bundled Parking Influence Travel Behavior? 2018 Pinski, Miriam Julia Advisor(s): Manville, Michael.

After controlling for differences in socioeconomic and built environment characteristics, I find that the presence of bundled parking is associated with a 27 percent increase in vehicle miles traveled. Bundled households drive approximately 3,800 miles more, spend nearly $580 more on gasoline, and emit 14.47 more metric tons of carbon dioxide per year. Bundled parking is also negatively correlated to transit use, and households with unbundled parking are significantly more likely to be frequent transit users. This provides further evidence for the already strong case against parking requirements.

Via.

[ related topics: Nature and environment Law Public Transportation ]

Today in great compiler errors, turns out that:

Argument of type 'Component<typename, unknown>' is not assignable to parameter of type 'string'.

Means I misspelled a member name in TypeName.

I loathe this language.

Inside Higher Ed: Brown Professor Suspects Majority of His Class Used AI to Cheat

As colleges and universities grapple with AI, cheating must be taken seriously, Serrano said. “We cannot afford to have a society in which a significant fraction of our best young minds think that cheating is OK,” he said. “That leads to a declining society, to a failed society … We cannot choose to become idiots.”

Via.

Brown University: Generative AI in Teaching and Learning (GAITL) Committee Final Report and Recommendations (PDF) was published while all of this was unfolding, and mentions that:

Among Brown student respondents, 56% of undergraduate respondents and 67% of graduate and medical student respondents reported intentionally using GenAI tools daily or weekly. Master’s degree students identified themselves as frequent users at the highest rate (85%), followed by medical students (77%) and then doctoral students (50%). Adoption of GenAI technologies also has distinct patterns across knowledge areas, with a large majority of students in the life sciences (79%) and physical sciences (73%) identifying as frequent users. Students studying the humanities and the arts had the lowest rate of frequent users (41%).

And then notes that the camel is already in the tent:

Google Gemini tools are currently approved by OIT for use by students and instructors, and are accessible, free of charge, through their Brown accounts. Subject to constraints due to cost and assurances around data privacy and accessibility standards, the University should provide access to additional GenAI tools, since they differ significantly in their strengths and weaknesses. For example, ChatGPT, Claude Code and Brisk’s quiz generator provide different functionalities and would be an excellent complement to Gemini. In addition, many services have “premium” licenses available, and there are equity issues if some students have sufficient resources to use them and others in the same classes do not.

I'm having trouble reconciling bemoaning the use of generative AI and LLMs to take your take-home tests for ya, while providing lie machines bundled in with tuition.

Edit: Tom's Hardware: Professor suspected AI-powered cheating on take-home midterms, makes finals in- person — only two students scored within 10% of their midterm score.

[ related topics: Interactive Drama Privacy Current Events Education Artificial Intelligence ]

Cheyenne Won’t Take Data Center Wastewater After Meta Contractor Contaminated System

Goat Systems LLC was in “significant noncompliance" with the city's industrial pretreatment regulations after discharging wastewater contaminated with Cupriavidus gilardii, a bacterium that interfered with operations at the city's water reclamation facilities and contaminated the municipal reuse water system, according to the BOPU’s Thursday statement.

Via.

Fuck yeah, let's gooooooo! Gizmodo: Meta’s Teen Safety Case Just Became a $1.4 Trillion Existential Threat

[ related topics: Law Pop Culture ]

Whoah. SIPEED Nano Home Lab: NanoXray. Okay, it's in development, not actually a shipping project, but:

Have you ever wished you had your own X-ray inspection tool? A compact desktop device for BGA checks, PCB analysis, and non-destructive inspection of small structures.

Via.