Mastodawn

Lots of AI-powered “news” sites make up bogus quotes, but making up at least six fake EFF sources in two months? That’s making a play for the championship title of bogus news content. https://www.eff.org/deeplinks/2026/06/news-site-keeps-hallucinating-eff-staffers

‘News’ Site Keeps Hallucinating EFF Staffers

What do EFF staffers Sarah Chen, Javier Morales, Caitlin Chin, Emma Rodriguez, and Mikko Kopponen have in common? For one thing, they don’t exist. For another, all have been quoted as EFF experts in articles published in the past two months on a site called News-USA Today, which describes itself as...

Electronic Frontier Foundation

Jeremy Kraybill 20h ago

If you can't define intelligent/conscious/sentient, you can't claim that something is or isn't those things you can't define. Ted Chiang's Atlantic piece was especially frustrating. https://www.theatlantic.com/philosophy/2026/06/no-artificial-intelligence-is-not-conscious/687378/

He also claimed in 2021 that computers will never self-improve: https://www.newyorker.com/culture/annals-of-inquiry/why-computers-wont-make-themselves-smarter

The hard problem of consciousness means we can't even prove that Ted Chiang is conscious. Confident dismissals of non-biological consciousness being ever possible are puzzling, esp from a sci-fi author.

No, Artificial Intelligence Is Not Conscious

Taken to its logical conclusion, this line of thinking is absurd—and damning.

The Atlantic

Show thread

Jeremy Kraybill 21h ago

@HistoPol it's a framework that lives inside Claude Code (or theoretically any other AI harness). It gives you projects where the AI behaves more like a trusted officemate than an unpaid intern. Claude Code "knows" when its context is about to be compressed, and under Project Gordo it usually suggests a fresh context instead of compressing (or when moving onto an unrelated task). It's the only way I use CC now.

Jeremy Kraybill 1d ago

Fable seems to be a bit more incisive/assertive of a reviewer than Opus: "This is one of the more intellectually honest documents of its kind I've encountered, and that honesty is also where its deepest problems become visible." Helpful, but ouch

Jeremy Kraybill 1d ago

Ars Technica 2d ago

Google's latest DiffusionGemma open AI model comes with a 4x speed boost

Diffusion AI is most common in image generation, but it can make text outputs much faster.
https://arstechnica.com/google/2026/06/googles-latest-diffusiongemma-open-ai-model-comes-with-a-4x-speed-boost/?utm_brand=arstechnica&utm_social-type=owned&utm_source=mastodon&utm_medium=social

Jeremy Kraybill 1d ago

Hacker News 2d ago

Anthropic's Model Naming, Extrapolated

https://samwilkinson.io/posts/2026-06-09-anthropics-model-naming-extrapolated

#HackerNews #Anthropic #Model #Naming #AI #Research #Machine #Learning #Technology

Anthropic's Model Naming, Extrapolated · Sam Wilkinson

With the release of Claude Fable, it is clear that Anthropic is progressing from poems to enterprise-scale narrative objects.

Show thread

Jeremy Kraybill 2d ago

@HistoPol I don't really have problems with hallucinations or with concentration / tokens. Not sure if the credit is to Project Gordo or Claude 4.5 Opus. Gordo resets its own sessions before it runs into ctx issues. If you use Claude Code, highly recommend trying Gordo Forge -- only about 15 mins to create a collaboration hub. It feels very different from openclaw etc -- check out https://github.com/gordo-ai, all the content there are "Gordo's spare time projects", no prompting from me.