It sounds as though they spend more time crafting the prompt/tool call that is sent to the image generator.

Be cautious when reading AI critics. They tend to be Western-chauvinistic in my experience. Even if every US-based AI company disappeared tomorrow morning, the Asian AI companies would continue dong fine. AI, big picture, is here forever. Some brash companies, like OpenAI, may not be.

In a nutshell, the ANES data shows:
📉 Social media use is shrinking; engagement collapsing
💥 Twitter/X posting has moved ~70 POINTS to the right
🧩 Platforms are splintering
🔊 Fewer people are talking — but those still talking are more politically extreme

Ain't nobody doing a massive port to Python. Target-rich orgs, like banks and EMR systems: legacy architecture and high-value data. They're going to (1) starve the monolith, and (2) Strangler Fig Rust, TypeScript or Go shims onto their C or Debian code. No Java or C#, too heavy for serverless.

Best I've got is the end of Carlini's [un]prompted talk. He says "in the limit, the defenders will win," but in the meantime things are scary. He sounded genuinely worried. He also predicted that maybe everything would be rewritten in memory-safe languages.

Unless you're a world class read teamer, it will be hard to know. An escalated attack might combine multiple vulnerabilities in a chain. One vuln that is "just a crash" might do much more when synergized. As long as they can be fixed quickly, it's better that they don't exist at all.

Edit buttons are security vulnerabilities, so, yes, Mythos could find that.

The fall of the theorem economy How AI could destroy mathematics and barely touch it

Great read

I for one welcome our theorem provers

I like to deal with concepts and intuition instead

Interfacing MCP with Combinatorial, Convex, and SMT Solvers - Stephen Diehl Personal blog of Stephen Diehl - Software engineer writing about technology, programming, and the future

This is really cool. MCP SAT solver for scientific computing. h/t @urschrei.eurosky.social www.stephendiehl.com/posts/smt_an...

prosaic

Not every book is worth finishing.

M'Gentleman.

The zero-days are numbered  | The Mozilla Blog Since February, the Firefox team has been working around the clock using frontier AI models to find and fix latent security vulnerabilities in the browser.

Report by Mozilla/Firefox about their security work with Mythos Preview. They believe the defenders will win, because it's finally possible for the defenders to find all the vulnerabilities as the attackers can find. blog.mozilla.org/en/firefox/a...

Pro-Iran hackers claim ‘sophisticated’ attack on Bluesky A pro-Iran hacking group that has been relentlessly targeting American companies with DDoS attacks claims it disrupted Bluesky with what the social media site called a “sophisticated” attack that…

We've made it, folks

Bluesky was down last week because it was hit by a geopolitical adversary — Iran, in this case

Screenshot showing poster jeremiahchronister labeled with "Made over 100 replies yesterday" and muted by the list "Reply/contact spammers, scammers and bad actors"

fyi

What gets me is, if, instead of infrasound, people talked about, "sound 9 year olds can still hear, but adults who have been living or working near it can't hear anymore," there might be much more cogent criticisms to level. No need to invoke the New Magics when the old industrialism will do.

A photo you posted (you probably know the one) got me lifting weights more regularly. So thank you for that.

meaning, "I built something for me, and I'm using it bc I'm invested. But how can I get others on my team, or on other teams, to use it too?" If you treat "suspicious of AI" as an accessibility factor, the role of the accessibility expert is clear IMO. Interfaces that meet people where they are.

I think this is a clear area where hiring should rubber band back. (Whether it will or not idk.) The AI/agent theme of 2026, at least among "thought leaders" is: getting agents to actually be productive in practical settings. And user interface/HCI is critical for that -- especially for MINDSHARE +

** 2006-10-31: The default prefix used to be "sqlite_". But then ** Mcafee started using SQLite in their anti-virus product and it ** started putting files with the "sqlite" name in the c:/temp folder. ** This annoyed many windows users. Those users would then do a ** Google search for "sqlite", find the telephone numbers of the ** developers and call to wake them up at night and complain. ** For this reason, the default name prefix is changed to be "sqlite" ** spelled backwards. So the temp files are still identified, but ** anybody smart enough to figure out the code is also likely smart ** enough to know that calling the developer will not help get rid ** of the file. */ #ifndef SQLITE_TEMP_FILE_PREFIX # define SQLITE_TEMP_FILE_PREFIX "etilqs_" #endif

Here is an insane piece of lore inside SQLite's source code

I am researching VACUUM and I was studying their code. In VACUUM, SQLite creates a temp file prefixed with `etilqs_`

Here is why:

A friend, a server at a high end steakhouse, once told me in utter seriousness, "With my PhD in Comparative Literature, I can get a job at any restaurant in the world."

4. An LLM made a mistake is a verboten sentence. You made a mistake: no offloading responsibility.

Some say there is no evidence infrasound causes harm to humans, and they are from diverse internet locations, like one written by an undergrad, never peer reviewed, only published to his personal web page.

The best example I have is from months ago, don't remember which model, a friend asked ChatGPT to write a poem titled "My Sweet Fetty," and ChatGPT one-shotted a poem about a recovering addict who was pining for fentanyl like a lover he ached for but knew was bad for him. It was good.

Genuine lol. Have a great evening.

^ fyi @den.dev

For my purposes, the thinking is clearly better than 4.6. But I'm seeing a lot (as in A LOT) of errors with the file system MCP. Writing to the wrong directory, realizing it and cleaning up, writing to weird places in the container instead of my fs. I wonder if that is the root of the complaints.

If you are using Google as a source, you could put this prompt into Gemini and relax: "What was the rate of electronic card catalog adoption in K-12 schools, once they were available to replace written card catalogs?"

I don't think that's controversial. Or, more correctly, I don't care whether it's controversial; it's a clear fact. Btw I use DuckDuckGo too, considering the switch to kagi.

It looks to me as though an LLM generated your citations and you ran with it. The references you use just don't relate to your claims. You aren't alone, by a long shot. Attorneys have done it in court. The takeaway I'd suggest for the future is to personally check all references before publishing.