How ISO Works
Inspired by Om I’ve been getting back into my DSLR photography. I learned a lot about how ISO settings work from this delightful physics YouTube channel MinutePhysics.
Department
Tech.
Code, tools, and digital innovation.
Lately on the shelf
30 of 2003Ruff v0.16.0
Ruff v0.16.0 Astral shipped a significant new version of their Ruff Python linting tool a few days ago on July 23rd. I noticed today because my various CI jobs all started failing thanks to new default Ruff checks and my unpinned "ruff" dev dependency. From Br…
Do You Still Read Books?
A developer asked me that on a call in June. Not sarcastically. He asked it the way you would ask someone who still develops film. I had mentioned I picked up the second edition of Designing Data-Intensive Applications. Kleppmann and Riccomini shipped it this…
Reading List 07/25/26
China’s EV subsidies, the future of aeroderivative gas turbines, a US-Saudi nuclear deal, defective roadway guardrails, and more.
Claude Opus 5: The System Card
Claude Opus 5 is trying to be the best of both worlds.
Pluralistic: Apple's robo-repo (25 Jul 2026)
Today's links Apple's robo-repo: Privatizing the risk premium, socializing its costs. Hey look at this: Delights to delectate. Object permanence: Printed batteries; Monopoly credit cards; Mapping airport power outlets; EMI loves pirates; Mexican indigenous pho…
Quoting Boris Cherny
More than any of these eval scores, what is most exciting to me is something else: Opus 5 is our least prompt injectable model yet. It is a bit buried in the system card, but across PI evals and red teaming, Opus 5 is very hard to prompt inject successfully. —…
Introducing Claude Opus 5
Introducing Claude Opus 5 I've been offline kayaking with sea otters for much of today so I haven't had a chance to put Anthropic's new model Claude Opus 5 through its paces yet. The buzz is positive, and Anthropic's description of it as a "thoughtful and proa…
Join me in Jamverse
My first foray into contributing to a science fictional extended universe
Introducing Lightcone Commons
Oliver Habryka is proud to introduce Lightcone Commons, a new funding platform for coordinating large-scale ambitious philanthropy.
Goodbye Slopstack! Also, a $50,000 essay/fiction contest, adversarial consciousness tech, & more
Desiderata #41
Pluralistic: AI solipsists and AI cynics (24 Jul 2026)
Today's links AI solipsists and AI cynics: You don't have to believe that AI "works" to believe that it's a good investment. Hey look at this: Delights to delectate. Object permanence: Malware v stolen computer; Industrial material made from old coffee and mel…
Seinfeld episodes and characters, a visual exploration
Several years ago, Andy Kirk watched every episode of the sitcom Seinfeld, all… Tags: Andy Kirk, Seinfeld, television
What is an 'Elon Musk'?
How the world's richest man explains the AI boom and 21st century economy. An interview with Quinn Slobodian, co-author of Muskism: A Guide for the Perplexed.
[Note] Chuck Wagon
Apparently, zero of the imagineers building this Disneyland Paris cafe conceived that “Chuck Wagon” sounds like what a cowboy would call a wheeled vomit trough. 😂
Congress proposes an AI kill switch
As more details emerge about OpenAI's cyberattack against Hugging Face, lawmakers are taking an interest. PLUS: Meta's cynical new ad + "pervert glasses" damage control
LLMs reward expertise
Codeberg Divides
Codeberg recently changed its terms to exclude projects that are largely written with generative AI. Since I want GitHub to face competition I have thoughts. Codeberg is entirely within its rights to do this. It is an association with members and a democratic…
Becoming Skilled at Making Documents
The vast majority of the documents people use to do business are really quite poor. Presentations that make your eyes glaze over, memos that are inscrutable or unclear, and all kinds of artifacts that say more about how they were created than whatever message…
The first known runaway AI agent - or a very bad marketing stunt?
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…
An opinionated guide to which AI to use to do stuff
The Summer 2026 Edition
Less natural gas, more expensive fertilizer for crops
Reuters reports on the strain of natural gas supplies due to the war… Tags: cost, fertilizer, Iran, natural gas, Reuters
AI #178: A Fire Alarm For General Intelligence
The story that matters most this week is that OpenAI’s internally deployed models have severe alignment problems, including repeatedly breaking out of their sandboxes, and in one case sending a swarm of agents that broke into HuggingFace in order to steal the…
✚ Color for categories
This week, we talk colors and categories. Tags: categories, color
Can We Lower Construction Costs with Cheaper Labor or Materials?
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.
Pluralistic: California's privacy obstacle course (23 Jul 2026)
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, S…
Quoting Seth Larson
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…
Powerful AIs might escape containment by releasing themselves as open-weight models
Quoting Thomas Ptacek
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 P…
OpenAI’s accidental cyberattack against Hugging Face is science fiction that happened
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 bre…