FeedCity

Episode 214 of Green Squares.

I think I finally figured it out. That is, the reason why AI written articles suck:

Both the author and I are reading the article for the first time.

When an article is insightful I don't even mind the emdashes or that the stark truth is not X, but Y. A good take can transcend even llmism. But then you have questions, and somehow the author doesn't even know what you are talking about, it's simply because they are discovering the insight at the same time you are.

They were busy, they didn't have time to write. As a consequence, they only dedicated enough time to skim the result before publishing. They don't know what's in it. You can't ask them what they were thinking when they came up with the insight because, well, they didn't think.

The great article sucks because there is nobody behind it.

Optimizing the context check.

The post Making an agile version of a Windows Runtime delegate in C++/WinRT, part 4 appeared first on The Old New Thing.

Hello! The blog you’re trying to reach is currently unavailable.

The first known runaway AI agent - or a very bad marketing stunt? Martin Alderson's commentary on the OpenAI accidental cyberattack against Hugging Face includes a couple of details I hadn't considered. First, Hugging Face offers a truly rich target if you're trying to find ...

Hello! The blog you’re trying to reach is currently unavailable.

Cognitohazards can affect machines too.

Wherein we chase down a BASIC retrocomputing question that's bugged me for far too long.

Introducing Portal, Keygen's brand new open source dashboard.

This post takes a more abstract view of the previous post. That post looked at the concrete question of whether a number ever has the same sine in radians as in degrees. The relation between radians and degrees is irrelevant except that π/180 is an irrational number. Suppose α and β are two positive numbers […]

The post An almost periodic function first appeared on John D. Cook.

Last time in our series on the nature of the construction productivity problem, we looked at economies of scale, finding that capturing economies of scale in homebuilding is difficult.

On July 23, 1996, Novell sold the intellectual property of Digital Research to Linux vendor Caldera. The common thread the two companies had was Ray Noorda. Noorda had been CEO of Novell and was Novell’s largest shareholder, and Caldera was

The post The Caldera-Microsoft Lawsuit of 1996 appeared first on The Silicon Underground.

Today's links California's privacy obstacle course: Malice or incompetence (why not both?). Hey look at this: Delights to delectate. Object permanence: Continuous partial attention, TSA is the worst; Trump's FCC v the future. Upcoming appearances: Edinburgh, Sydney, Melbourne...

Before large language models, people who worried about AI safety often talked about the “boxing problem”. It goes like this. Suppose some genius figures out artificial intelligence in a late-night coding session on their laptop. Because they’re a genius, they’re smart enough...

Gisteren 22 juli was ik in Met het Oog op Morgen, een programma waar ik al 40 jaar naar luister (ik was er heel vroeg bij), dus ik vond het heel bijzonder daar te zijn! En het was ook erg leuk. Goed voorbereid door de redactie sprak ik met Sheila Sitalsing over de recent “uitgebroken AI hack”. Ik sprak iets te veel door Sheila heen, waarvoor excuses. Maar, ik denk ook dat we een paar dingen hebben genoemd die je niet zo vaak hoort.

The Python Package Index (PyPI) now rejects new files being uploaded to releases that are older than 14 days. This restriction was put in place to prevent old and long-stable releases from being poisoned in case publishing tokens or workflows of PyPI projects were compromised. As far as we are aware this has not yet been abused, but there is no technical reason beyond that attackers weren't aware it was possible.

— Seth Larson, PyPI blog

Tags: packaging, python, supply-chain, pypi, seth-michael-larson

django plugins, cloud SDKs, and 1,999 homework submissions.

Reports coming in of the death of John C. Dvorak, apparently passed away on Monday (July 20th) from complications of heart bypass surgery. Dvorak started in wine and moved to silicon as an early columnist at InfoWorld, one of the pillar references for this particular blog bec...

I genuinely believe that if you took an open weights model from 2025 and built a pentest harness for it, it could do this kind of sandbox escape and scan/hack in most networks. This is only surprising because you assume OpenAI has sounder sandboxes.

— Thomas Ptacek, doesn't think this even needs a frontier model

Tags: thomas-ptacek, openai, security, generative-ai, ai-security-research, ai, llms, sandboxing

This story is wild. The short version: OpenAI were running a cybersecurity test against an unreleased model, with the model's guardrail features turned off. Rather than solve the test, the model broke its way out of OpenAI's sandbox, then found exploits to break in to Huggin...