A Redistribution of Code Ownership
Scott's Blog
Scott's Blog
The price of synthetic labor is collapsing. Some thoughts from my small agent factory about AI economics, KV cache compression, human-in-the-loop and whether I should feel guilty about keeping the machines busy.
Jev is a useful zero-shot classifier, but its probabilities can't be calibrated for your data. Calibration depends on your data distribution, which Jev never sees, so treat its outputs as scores and recalibrate them on your own labeled examples.
The first time AI makes me tear up I'm biking down a sunny Brooklyn street.
Jobs of the Future May Resemble Those From the Past
The jobs of the far future may not be ones we can’t conceive of yet, but instead ones from the past that keep us capable of operating independently of AI. As recent geopolitical events can attest to, there is a risk in trusting partners with critical capabilit…
10-year Treasury yield Recently, interest rates for government debt have been rising worldwide. As I write the yield of the 10-year US trea...
A benchmark where frontier language models drive a real comma-equipped Toyota through a cone course, one command at a time, with a human supervisor ready to brake.
A message can arrive on time and still belong to a world that has already changed. What a paper from 1978 can teach us about group chats, shared decisions, and the computers we trust to act for us.
– Children playing ball games, Roman artwork, 2nd century AD. Photo by: Jastrow via Wikimedia Commons, CC BY 3.0 One of the best laughs I’ve had with my wife have been due to sharing childhood experiences. She’s also fond of my baby memories, or the really emb…
Thank me, I just listened to a 5 hour interview with DHH so you don’t have to. I would sum it up as a never-ending serenade to the current love...
I work on a service that decides which courier gets offered your delivery at a gig economy delivery app. When you open the app and ask for something to be brought to you, some code somewhere has to decide which of the couriers should be offered the job. We’re …
BrewUI is Homebrew's official macOS GUI. I'm a terminal person by default, but here's why I installed it anyway, how the configuration works, and what upgrading and installing packages looks like from the app instead of a shell.
A long time ago I ran a write-heavy system on a hub and a handful of workers. Each worker took a share of the application traffic and wrote events locally. The …
Case study on Germany’s exit tax: how the author spent 1.5 years, consulted 13 tax advisors, and paid 18.7k€ to move to Thailand.
To my knowledge, the first recorded LLM-agent NetHack ascension
In early results from our new life sciences research lab, Claude agents found an enzyme system whose function is still unknown.
Jev gave a hidden fair die roll an 83% probability and was right 19% of the time. A check of its calibrated probabilities, with code and data.
what belongs in a research harness when the models keep improving, and how do we get the expertise into it?
The other day I merged a new feature for Inko that I think is quite interesting: the ability to sandbox an application with minimal effort.
In most of the cases you don't need (Astra|Fable|Opus|%paste-a-new-LLM-name-released-this-week%).
On first principles thinking, learning to work with agents, and setting experience aside long enough to see a problem clearly.
Coupling and Cohesion in practice: meaningful quantitative metrics and the common ground behind software design principles.
Log in or sign up to personalize your feed, join conversations, vote, and explore communities.
Claude Code 2.1.277 added AGENTS.md support, but the loader sits behind a remote feature flag. With telemetry or nonessential traffic turned off, a local AGENTS.md is skipped without a warning. This is what I measured and the one-line CLAUDE.md I use instead.
Everyone is talking about Jev - here it is in 25 lines of Python.
SAML, the XML-based authentication protocol that birthed the SSO industry, is fundamentally flawed due to XML complexity, canonicalization issues, enveloped signatures, and design ossification, making it vulnerable to signature wrapping attacks and parser diff…
FoxDev Studio
A modern IDE and 64-bit runtime for Visual FoxPro 9 applications: the same language, rebuilt on Electron, React and a Rust virtual machine in WebAssembly.
Analysis of Anthropic's Claude Opus 5.5 (Adaptive Reasoning, Max Effort, Default Fallback) and comparison to other AI models across key metrics including quality, price, performance (tokens per second & time to first token), context window & more.
The author should be the first human to review a change, before asking someone else to approve it.
Superhuman programs won't teach you to be superhuman.
Introducing Oh My Quant: a research-native agent harness for experiments, durable evidence, and traceable conclusions.
As one who’s lived in the shell for twenty years, I’m happy (though not surprised) to see how naturally the CLI fit into the AI world.
Hardware gets cheaper, general models handle robot tasks, and full-stack robotics companies reach real revenue. What should others in the space build?
For some inexplicable reason, macOS won’t let you turn off the MacBook’s built-in display when it’s connected to an external monitor. The free One Display utility fixes that.
tokens are going to be as cheap as electricity within the decade
The scheduler must live at the same level as the constrained resource.
Reference for early career folks asking for help
Be considerate of the time you require from other community members.
If your system only survives you sober, it doesn't survive you.
What I use to run large language models on self-hosted GPU on AWS, and how it cut my Claude bill.
Roman Emperors, blog writers and super heroes all have to find the values they live by somewhere - Marcus Aurelius, Mark Manson and Oliver Queen (The Arrow) tell us where.
I Bought A Scanner (No, Really This Time)
My grandfather was a plumber. Correction: my grandfather was the best damn plumber in his North Jersey suburb of New York City, and everyone in town knew it. He had ten kids, six of them boys, and as my father and his five brothers grew up they all, at one poi…
Introducing Validator, a CLI for evaluating Jev model outcomes.
The preface to my PhD thesis — on AI, how research is changing, and the kind of researcher I want to become.
@tranquil_cassowary@infosec.exchange @ocelot221@mastodon.social @mhoye@cosocial.ca @mulchmaxxing@blahaj.zone There's a high chance of devices being sold with it preinstalled in 2027 but probably not for the initial launch.
Try adding an item to each list below.
Your org is the quality group from the pottery parable. They were told to make one perfect pot. They made worse pots.