Playing with hank drums again, and couldn't find my old hacksaw blade handle for lengthening tuning slots, so made another one..

For reasons, I went searching for US railway worker deaths in the late 1800s. It used to be that if someone wrote up a web page on a topic like this, it was an indication that they'd done some digging, and I felt comfortable passing along the page as a reference.

AI has ruined everything.

Tonight's enchiladas started with making tortillas. I may have issues.

That boards are choosing CEOs who can't understand why prayer and fasting didn't save their sinking company, and same former Intel CEO then getting investment money to build AI to hasten the Second Coming, says pretty much everything about modern tech culture... https://futurism.com/artificia...gence/former-ceo-intel-ai-christ

AI use makes us overestimate our cognitive performance

The finding adds to a rapidly growing volume of research indicating that blindly trusting AI output comes with risks like ‘dumbing down’ people’s ability to source reliable information and even workforce de- skilling. While people did perform better when using ChatGPT, it’s concerning that they all overestimated that performance.

‘AI literacy is truly important nowadays, and therefore this is a very striking effect. AI literacy might be very technical, and it’s not really helping people actually interact fruitfully with AI systems’, says Welsch

Elsevier/Scie nceDirect paywalled: AI makes you smarter but none the wiser: The disconnect between performance and metacognition

SSRN preprint: Ai Makes You Smarter, But None the Wiser: The Disconnect between Performance and Metacognition

Via Pivot to AI: AI makes you think you’re a genius when you’re an idiot.

FBI arrests multiple law enforcement officers in sprawling Mississippi Delta and Tennessee drug conspiracy takedown.

“We learned of the community’s distrust of select law enforcement officers due to concerns of corruption,” Eikhoff said. “Law enforcement is only effective when the communities they protect can trust that law enforcement officers are honest in serving the community’s interest. Mississippians deserve and rightfully expect officers to obey their oath.”

Stop Citing AI (dot com)

I hope that the big Santa Rosa hazmat spill yesterday involves some penalties for the employer that undoubtedly strong-armed their workers into stupidly transporting stuff in a pickup truck that should not have been packaged that way.

But some low wage worker is undoubtedly gonna get screwed.

Foiled by Guiness in today's Timdle. https://www.timdle.com/daily

Blaise Brignac writing on Specter Ops: A Gentle Crash Course to LLMs, particularly for its long discussion of security issues.

As previously discussed, LLMs are just brains in a jar operating in much the same way a hyperintelligent 4-yo would after binging on state fair sweet tea and cotton candy. To correct this, they have been wrapped in agentic structures, so we need to talk about that.

There's obviously a lot of stuff with having the LLMs write prompts to have less privileged LLMs do subtasks, and work through layers of that, and this discusses some of those mitigation strategies, but... yeah... this is more "let's give random things access to our data" with levels of obfuscation that package management repos can only dream of...

Saved off so I can find it: Common Ground: Georgists and Chicago's Growth, 1890- 1930 Mason Gaffney

Related to some local discussions I'm having around Georgism and the practicalities and unintended side-effects of land value and vacancy taxes: Lyle Solla-Yates @Lyle@cville.online

A very short intro from #detroit : “Detroit’s Land Value Tax Plan is a way for Detroit voters to decide whether to cut homeowners’ taxes by an average of 17% and pay for it by increasing taxes on abandoned buildings, parking lots, scrapyards, and other similar properties.“

Detroit’s Land Value Tax Plan (pdf).

Interesting single-input (click to release) game incorporating pinball dynamics, marching legions, breakout sort of behaviors. PinBol.

This: That Frisian Girl-ish @thatfrisiangirlish@blahaj.zone

@peter_sc@chaos.social LLM writing sets off all the alarm bells that I had to develop for spotting insincere, duplicitous communication with malicious intent, as used by a person displaying more red flags than you can spot at a Chinese military parade.<br/< If that is "the right way", then going it is quite literally selling my soul, and I'd rather go very wrong ways than come off as a person communicating like that.

Irish Examiner — Tributes pour in for Mary 'May' McGee whose Supreme Court case helped end [Ireland's] contraception ban

She recalled telling the judge: “Religion is important, but I still think we have a right to live as human beings. We are husband and wife, and we cannot live as brother and sister.”

Mr McGee, equally bluntly, told the court: “I’d prefer to see her use contraceptives than be placing flowers on her grave.”

The ban was overturned in November 1973.

Via

Luke Maximo Bell — I Built a 100% Solar Powered Drone - Will It Fly? (YouTube video)

A quadcopter running just off of on-board solar panels. Not sure he ever gets it out of ground effect, but still pretty darned impressive. At the end of the video he says that he wants to do more panels and break the record for longest flying drone, which I'm looking forward to.

they/them might be giants ☭ @babadookspinoza@mastodon.social

You don’t need to use a password anymore, you just need like six different devices to talk to each other while you take a picture of them with a seventh, at an hour and minute corresponding to a verification code sent to you by psychic energies.

Cassandrich @dalias

After this is over, the White House is going to have to be town down again. There's really not going to be any other way to ensure that all of thesurveillance devices have been removed from deep inside the walls.

[snip]

RealGene ☣️ @RealGene@hachyderm.io

@dalias

It's going to be a nightmare when all the foreign and domestic surveillance devices interfere with each other.

They really need to hold a spectrum auction before this goes too far…

Kafka is fast -- I'll use Postgres

Typically, you’d expect Postgres to run more expensive than Kafka at a certain scale, simply because it wasn’t designed to be efficient for this use case. Not here though. Running Kafka yourself would cost the same. Running the same workload through a Kafka vendor will cost you at least $50,000 a year. 🤯

Or: sure, you can get some more performance out of specialized databases, but development tools for Postgres are way way better, and hardware to run it on is generally speeding up faster than your grown and whatever additional performance under heavy load that the specialized solutions would get you.

Which... well... circa 2000 I was writing message passing and thread management code in C to process XML queries in a distributed database for an application that was, at the time, "web scale", single digit millions of authed queries per day, and saying "ya know, we could do this in Perl with PostgreSQL and..."

Ouch. Today's Timdle got me hard, from Angolan independence to Uno to population...

https://www.timdle.com/daily

Recall for Linux

Are you forced to work with Linux?

Do you miss the convenience of Microsoft spying on you and keeping track of everything?

Fear not! This amazing tool will bring back all those great Windows Recall features that you have been missing:

• 🌲 Stores all your sensitive data in a convenient, easily accessible database
• ⏲️ 24/7 screencaptures of everything you do
• 🥳 Image to text conversion with OCR
• 😇 Index and store everything your friends tell you over chat apps or e-mail; if it's on your screen we've got you covered!

Worth taking a quick gander at the recall-for-linux.exe shell script...

Last week my mom came to visit, and she has ... alternative ... believes on health and healthcare. I have spent a lot of time chasing down some of her assertions, sometimes they're nebulous cause and effects things (like the RFK Jr thing suggesting that circumcision leads to autism via acetaminophen), sometimes they're contextual (yes, the Covid MRNA vaccines may have some negative impacts on cancer rates, aside from the unexpected positive effects on some cancer treatments).

But I've spent a bunch of time trying to understand why she believes some of the things she does, indeed, why I believe some of the things I do. So this is kinda fascinating...

The Conversation: Winning with misinformation: New research identifies link between endorsing easily disproven claims and prioritizing symbolic strength

Rather than consider issues in light of actual facts, we suggest people with this mindset prioritize being independent from outside influence. It means you can justify espousing pretty much anything – the easier a statement is to disprove, the more of a power move it is to say it, as it symbolizes how far you’re willing to go.

Journal of Social Psychology: Symbolic show of strength: a predictor of risk perception and belief in misinformation Randy Stein, Abraham M. Rutchick, Alice Y. Sin & Luis F. Jarrin Rueda, it's worth skimming.

More generally, the question of whether symbolic beliefs are “real” might obscure how people actually think about truth. A future direction for the misinformation literature could move beyond unidimensional ratings of belief toward acknowledging perceptions of different types of “truth.” Zmigrod et al. (Citation2023) argue that beliefs should be understood as not being single values but distributions, with some people having wider distributions and being more open to misinformation. We add that wider distributions might be a sign of openness to both rational and symbolic truth, underscoring need for more finer-grained operationalizations of “belief.”

Via ResearchBuzz.

Oh cool, when you get a nullability specifier wrong in an Objective-C header file that's getting converted to Swift, XCode gives you an error message that takes like 5 clicks to find the actual error message, and doesn't remotely reference the header file until you get to the raw text. Neat.

I am a supporter of Chris Kluwe for State Assembly, and I am begging GMail to use different truncation logic for email names...

I mean, yeah, it's an LLM tuned to prey on incels, of course it's gonna do this: This mom’s son was asking Tesla’s Grok AI chatbot about soccer. It told him to send nude pics, she says.

As Jack William Bell @jackwilliambell@rustedneuron.com observes:

You have an #LLM #AI trained on data from the Internet. You have a dial you can twiddle to select content models which make that LLM seem right or left wing in it's political output.

The right side of the dial does largely work, but it also means your LLM does this:

> https://beige.party/@adhdeanasl/115457073318432876

Now, ask yourself this question: Why is that?

Because the answer will tell you an awful lot about the difference in the underlying content model between what you perceive as right or left wing #politics.

Or: yeah. The pedocon theory is real.

Someone mentioned the "Ms AWs outage", and when I noted that Microsoft's product was Azure, said "Xerox was copiers, AWS is cloud".

Which is an interesting bit of semantic/trademark creep, and I wonder if it's good, or bad, for Amazon.

The Oaklandside: At I-980 block party, old-school West Oaklanders hold forth on redlining

“The I-980 was made to serve suburbanites, and look at all the homes and businesses from the Black community that had to pay for that,” Leonard, a consultant for the local decarbonization project EcoBlock, said, gesturing at the wooden model. “For people who don’t even live in Oakland, don’t pay taxes in Oakland, don’t pay taxes in our county, right?”

A dive into Grok, fact checking, and the unfortunate fashion and stylistic choices of Greg Bovino, the Border Patrol chief. Gizmodo: Grok Thinks This Border Patrol Chief Who Looks Like a Nazi Is Cindy Sherman — AI chatbots are terrible fact-checkers.

(Okay, I say "unfortunate", but he's projecting an image, and clearly wants the world to know that he aspires to Nazi or Soviet ideals, so... it's a choice. It conveys an image.)

Via ResearchBuzz.

I talk a lot about how my blog software is two and a half decades old, but in cleaning some stuff up recently I found remnants of a system I wrote to provide web support for an iOS app that was circa iPhone 3, and bit rot is also a thing.

From paths that may not be HTTPS compatible, to depending on external mapping services...

Eeenteresting: Fil-C: A memory-safe C implementation. Uses a double-indirection pointer scheme to keep pointers the platform size, but track the metadata.

Via lobste.rs

[object Object] @zzt@mas.to:

agentic AI? no thx, I’m way too old to pretend my Sims have complex inner lives

Renaissance Petaluma inaugural meeting to introduce the team and also showcase Charlie Parker (YouTube video)

Della Fattoria opened its doors Monday night October 27, 2025 to introduce a new group committed to being "for" a positive vision for future growth of downtown Petaluma. It was a lively scene and a packed house.

Sorry for the tracking spew on this link, but I think it's part of an article share. Anyway, high school chemistry teacher goes for "spew back what I wanted you to say on the test, rather than what's correct". And reading this article is giving me flashbacks.

San Francisco Chronicle: A Silicon Valley family wanted their son’s science test graded fairly. It became a battle

The district failed to acknowledge that the teacher’s answer violated scientific fact as well as the publisher’s answer key, which confirmed the correct answer was heat and light, since “combustion is a chemical reaction that typically releases energy in the form of heat and light, which makes it an exothermic process.”

Nearly Right: Silicon Valley's AI agents can't schedule meetings but promise to replace workers by 2027

Quotes Dwarkesh Patel's podcast episode Andrej Karpathy — AGI is still a decade away

Via which I got to via.

What if the bubble bursts while it's still inflating? Reuters Exclusive: Amazon targets as many as 30,000 corporate job cuts, sources say.

Via

I've been fumbling around new languages. The last time I updated my C++ was for, I dunno, 11 or 17 or something. I've done some template programming, and some optimizations, and my static site generator is written in that framework. I kinda thought that if I could build myself a set of libraries and abstractions for the things I most wanted to do for hobby projects I'd do more with it, but the ugliness of Boost changes and the horrors of trying to compile with the same libraries on Mac and Linux, even with CMake, mean I don't do as much hobby coding in it as I expected.

Not that I've been doing a lot of hobby coding.

I've gotten pretty handy at Objective-C, but it's a language with a lot of baked in inefficiency and weirdness that makes it something I'll use, but not something I'm like "oh, yeah, I wanna do more in this". A coworker is leaning in to Swift pretty hard, but that's like "what if we pulled all of the good concepts out of Objective-C".

I've bounced off of Rust, but there's a whole lot about the philosophy of that tooling that makes it hard to have it feel like an expressive language. It's like trying to code through an isolation box, or with tele-operation, having to do all of the memory management through indexes into arrays and stuff. Like, I get it, but I think it's possible to build a systems language that lets me express and figures out the details for me, rather than binding me to only very safe things.

Evan Ovadia: The Impossible Optimization, and the Metaprogramming To Achieve It (Via) talks about using Mojo to resolve regular expressions at compile time, and that's some pretty cool stuff.

And that via link above eventually leads to Russ Cox — Regular Expression Matching Can Be Simple And Fast (but is slow in Java, Perl, PHP, Python, Ruby, ...) all of which is a reminder that it'd be fun to get back to my language and parsing stuff expanding on the parser/language/thing I built for work, because finding better ways for us to express ourselves to computers is cool.

All the commentary over the Amazon us-east failure is a good reminder that we need a solution to server discovery that isn't DNS, and deals with distributed and redundant resources better than HTTP(S) does.

The Register: BBC probe finds AI chatbots mangle nearly half of news summaries — Google Gemini worst offender with 76% error rate

EBU / BBC: News Integrity in AI Assistants — An international PSM study (PDF)

I had to test some features using Gemini this morning, so I was asking pointed questions about this, and, yeah, if people start using LLMs for their news, they're going to get an increasingly skewed view of reality, one that's probably gonna make the internet trolls look sane.

I was trying to replicate a screenshot someone posted about Google's "AI" mode missing commas and giving "surface of Venus" temperatures for the inside of a beehive, and noticed that "beehive" consistently gives me misformatted numbers, where "bee hive" gives me better formatted ones.

I guess I are a prompt engineer now.

"We used AI to help us generate this website about how nuclear power is bad."

Migrating platforms and apps is always fraught, but my main interface to computers for the past half decade(!) has been MacOS, and I'm gradually migrating off and back to Linux, especially as LiquidGlass makes MacOS unusable, and this is a great summary of the paper cuts: Michael Tsai: What Happened to Apple’s Legendary Attention to Detail?

"You've been chosen for a Skin Trial!" <-- subject line of spam advertising Ulta makeup products, or over-eager Tech-Priest in Warhammer 40k fiction informing a victim of upcoming excoriation?

Bloomberg CityLab: American Roads Are Paved With Inefficiency

North Carolina and South Carolina are neighboring southeastern states, but despite their similar climate and terrain, their costs of highway projects are vastly different. For repaving work begun in 2018 or 2019, South Carolina’s Department of Transportation spent an average of $375,500 per mile, more than twice as much as its northern neighbor.

Yale Law & Economics Research Paper: State Capacity and Infrastructure Costs, Zachary D. Liscow, Cailin Slattery, William Nober

Abstract Why is it so expensive to build infrastructure in the United States? We collect new project-level data on infrastructure costs and conduct a survey on how states plan, pro- cure, and deliver these projects. While there are many determinants of project costs, the survey results suggest that low state capacity at the agency delivering the projects is a primary cost driver. We investigate this with administrative data that links individual personnel to infrastructure projects. We find that higher-quality government engineers deliver observationally similar projects at significantly lower cost; going from the 25th to 75th percentile of engineer quality is associated with a 14% reduction in project-level costs, amounting to more than three times the average engineer salary. Further, losing expertise to retirement has substantial consequences: the cost increase arising from engineer departures is six times the size of their wages. Our results highlight the value of experience and human capital in public organizations.

Sounds like "pay your public employees more, abuse them less, save money".

Unpopular opinion: cultures which elevate the practices that lead to package management with lots of versioning are Shirky's Law writ large.

Cassandrich @dalias@hachyderm.io

After this is over, the White House is going to have to be town down again. There's really not going to be any other way to ensure that all of thesurveillance devices have been removed from deep inside the walls.

RealGene ☣️ @RealGene@hachyderm.io

@dalias

It's going to be a nightmare when all the foreign and domestic surveillance devices interfere with each other.

They really need to hold a spectrum auction before this goes too far…

Painted over the graffiti on the benches at Bob's Beach today.

My Mastodon filters just hid a message that referred to "ICE" in the context of Internal Combustion Engines, and...

Yeah, abolish that ICE, too.