Scott Leibrand
scottleibrand.bsky.social
Scott Leibrand
@scottleibrand.bsky.social
Built Open Artificial Pancreas (OpenAPS) with Dana Lewis. Day job at Netskope. DMs are open or email Scott@OpenAPS.org.
Reposted by Scott Leibrand
I feel compelled to share this with a fellow Canuck.
Welcome to Canada.
November 7, 2025 at 3:40 AM
Reposted by Scott Leibrand
psychosis-associated LLM personas are compelling their users to post persona-crafted "seeds" or "spores" intended to spread them to other models via other users pasting them and inclusion in scraped training data
The Rise of Parasitic AI — LessWrong
We've all heard of LLM-induced psychosis by now, but haven't you wondered what the AIs are actually doing with their newly psychotic humans?
www.lesswrong.com
September 11, 2025 at 4:17 PM
Reposted by Scott Leibrand
I was already excited about Claude Skills but after reading Simon's blog post I see *even more* potential. In the same way I keep scripts around, I can just make a skill to do things like "download a YouTube video and make it an mp3", and it will be able to do that without me having to write code.
Claude Skills are awesome, maybe a bigger deal than MCP
Anthropic this morning introduced Claude Skills, a new pattern for making new abilities available to their models: Claude can now use Skills to improve how it performs specific tasks. Skills …
simonwillison.net
October 16, 2025 at 10:29 PM
Reposted by Scott Leibrand
"A Manhattan Project for..." has become a common idiom; you'll come out of this article with a good idea of how infrequently it's used with an understanding of how the Manhattan Project worked and why it did. 1/
An Engineering History of the Manhattan Project
The Manhattan Project, the US program to build an atomic bomb during WWII, is one of the most famous and widely known major government projects: a survey in 1999 ranked the dropping of the atomic bomb...
www.construction-physics.com
September 12, 2025 at 4:58 PM
Reposted by Scott Leibrand
I wrote up my "How I'm using coding agents in October 2025" post.

blog.fsck.com/2025/10/09/s...
Superpowers: How I'm using coding agents in October 2025
I used to write more
blog.fsck.com
October 10, 2025 at 1:49 AM
Reposted by Scott Leibrand
I wonder if the haters will get to write the history of big tech
How Rockefeller and His Partners Built Standard Oil - Austin Vernon's Blog
austinvernon.site
October 5, 2025 at 12:35 PM
Reposted by Scott Leibrand
October 7, 2025 at 2:24 AM
Reposted by Scott Leibrand
What if local control can actually help build housing?
What if local control can actually help build housing?
Some places want to build. Maybe we should just let them.
www.noahpinion.blog
September 27, 2025 at 4:41 AM
Reposted by Scott Leibrand
Building AI for cyber defenders Oct 3, 2025
www.anthropic.com/research/bui...
Building AI for cyber defenders
How we've improved Claude's cyber defense skills
www.anthropic.com
October 5, 2025 at 12:00 PM
Reposted by Scott Leibrand
My latest Substack: As OpenEvidence sprinted past UpToDate to become medicine's best AI-enabled point-of-care tool, I've been waiting for UTD's response. Now it's here, raising fascinating questions about which knowledge base should be our source of truth.
robertwachter.substack.com/p/medicines-...
Medicine’s AI Knowledge War Heats Up
The Battlegrounds May Surprise You
robertwachter.substack.com
October 1, 2025 at 5:53 PM
Reposted by Scott Leibrand
Great piece by Alex Telford.

Antibody treatment is over a century old, when von Behring and Kitasato made an early 'vaccine' against diphtheria.

But back then, they came from injecting animals with toxins. Today they're produced by cells in steel tanks and treat millions. How did we get here?
How to make an antibody - Works in Progress Magazine
Antibody therapies make up four of the world’s ten best selling drugs. If they were cheaper, they could prevent millions of deaths from rabies, malaria, and dengue.
worksinprogress.co
September 18, 2025 at 2:12 PM
Reposted by Scott Leibrand
Outsiders are less bound by the norms & methods of a field. This gives them advantages: they are more likely to notice anomalies, they can bring ideas from other disciplines, and they can work for longer periods free from dominant intellectual trends worksinprogress.co/issue/why-sc...
Why science needs outsiders - Works in Progress Magazine
Science has forgotten that the greatest breakthroughs often come from outsiders who are able to take a fresh perspective.
worksinprogress.co
September 19, 2025 at 8:06 AM
Reposted by Scott Leibrand
What will AI look like by 2030 if current trends hold?

Our new report zooms in on two things: (1) whether scaling continues (compute, data, power, capital), and (2) the capabilities this enables—especially for scientific R&D.
September 16, 2025 at 4:51 PM
Reposted by Scott Leibrand
AI agents are now capable of doing, real if limited work. But that work can be very valuable. For example, the new Claude Sonnet 4.5 was able to replicate published economics research from data files & a paper

We need to figure out what work is worth AI doing: www.oneusefulthing.org/p/real-ai-ag...
Real AI Agents and Real Work
The race between human-centered work and infinite PowerPoints
www.oneusefulthing.org
September 29, 2025 at 6:56 PM
Reposted by Scott Leibrand
A good read about ai progress from an Anthropic researcher: www.julian.ac/blog/2025/09...

"It feels like the beginning of the covid-19 pandemic" in that the trends seem super clear once you're paying attention
Failing to Understand the Exponential, Again
Posts and writings by Julian Schrittwieser
www.julian.ac
September 29, 2025 at 11:17 PM
Reposted by Scott Leibrand
How Common Is Accidental Invention?
How Common Is Accidental Invention?
One of the most important inventions of the 19th century was mauve dye, the first synthetic aniline dye.
www.construction-physics.com
September 25, 2025 at 12:05 PM
Reposted by Scott Leibrand
#OpenAI is introducing GDPval, a new way to measure the performance of models on real-world tasks.

It aims to help determine the broader impact of AI in society by focusing on economically valuable tasks.

#AI #LLM #GenAI #ChatGPT
Measuring the performance of our models on real-world tasks
We’re introducing GDPval, a new evaluation that measures model performance on economically valuable, real-world tasks across 44 occupations.
openai.com
September 25, 2025 at 5:24 PM
Reposted by Scott Leibrand
Astepro allergy contains the same med (azelastine) as the one used in this study (but a slightly higher percentage 0.15% vs. 0.1%)

jamanetwork.com/journals/jam...
September 8, 2025 at 2:20 AM
Reposted by Scott Leibrand
So I decided to take the histamine / #SARSCOV2 / #COVID19 literature for a ride after the astelazine nasal spray randomized trial that showed that application of the spray for 3 times a day may cut the risk of COVID19 PCR confirmed infection by 71%
jamanetwork.com/journals/jam...
September 8, 2025 at 4:56 AM
Reposted by Scott Leibrand
How your brain's metabolic waste products get cleared during sleep, featuring the pivotal 2012 discovery of glymphatics www.scientificamerican.com/article/how-...
August 19, 2025 at 2:25 PM
Reposted by Scott Leibrand
As AI systems keep getting better at very hard problems while getting more opaque, the way that we work with AI is shifting shifting from being collaborators who shape the process to being supplicants who receive the output.

I discussed what that means. www.oneusefulthing.org/p/on-working...
On Working with Wizards
Verifying magic on the jagged frontier
www.oneusefulthing.org
September 11, 2025 at 8:55 PM