Harness design for long-running application development Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.

Oh hey this is just about the same as dolls methodology.

www.anthropic.com/engineering/...

Check your spam emails folder weekly! I need to remind myself often. One of my target conference acceptance emails went to spam with reviews that were as good as they could get and here I was wondering for a week why they hadn't gotten back to me.

How a clean heat transition could be delivered Exploring the market opportunity for Plymouth’s supply chain

A big update on Nesta's clean heat neighbourhoods work this morning: we've published a version the local heat plan we've been working on with Plymouth…

Fun one today

tired: thinking about tokens and context
wired: thinking about state and harness

GPT 4.1 for which i seemingly have unlimited quota in github copilot agent mode is a dumb model. Much less intelligent compared to sonnet 4.5 which is my go to model

Restarting my mathacademy journey. Biggest unlock in aggregating those xps daily has been using my phone instead of computer. Load in Firefox and note app with the touch pen for rough scratches.

your startup could have been a SKILL.md

Subscribe to Volts A newsletter, podcast, & community focused on the technology, politics, and policy of decarbonization. In your inbox once or twice a week. Click to read Volts, a Substack publication with tens of thou...

It seems like a good time to remind folks that Volts is 100% listener supported. My only obligation is to you. Subscribe & help me keep doing it!

Nice seeing you here. You are one of the names I remember from Twitter whose posts I always found valuable

I wrote about how I don’t know math but still am somehow a successful computer scientist. I have strong feelings about this. But I also want to understand.
togelius.blogspot.com/2026/02/math...

Some relatable feelings there. While I have been able to do derivations and "math" as needed, although with various levels of struggle, visual understanding is what moves the needle for me too. Unless my visual understanding and intuition is sharp enough, I feel like I haven't understood it well.

Create an agentic system to support my literature search

Write an article about it on my newsletter

Anytime a paper seems really relevant, to the point it will meaningfully improve your understanding or results, summarize the article and write how you might use lessons from the paper. Engaging with AI such as Claude to improve your understanding is very helpful.

LinkedIn is a great source of finding papers from people (connections) doing research close to your area of interest. Utilize opportunities for meaningful connections through comments, DM or sharing your own posts. No AI slop.

Notes to self on keeping up with the latest AI/applied research to improve my work:
- Bookmark anytime you come across an interesting paper/blog/article. Organizing all bookmarks in a single page is a must. Obsidian probably the best way to do it.

halftime show fromt eh supebowl showing a set on the field featuring a nyc bodega

This kind of smart, walkable, mixed-use urbanism is illegal to build in most American cities.

Bluesky Map Interactive map of 3.4 million Bluesky users, visualised by their follower pattern.

I made a map of 3.4 million Bluesky users - see if you can find yourself!

bluesky-map.theo.io

I've seen some similar projects, but IMO this seems to better capture some of the fine-grained detail

Wonder Man has heart and soul, nice change of pace and dorection from other Marvel series. The poetry-off was deep

Climate change risk index and municipal bond disclosures of United States drinking water utilities - Communications Earth & Environment 67 million customers across the US rely on drinking water utilities that face higher climate risk than accounted for, which exposes major gaps in climate adaptation and resilience planning, suggests a...

🚨Our new paper led by Dr. @zialyle.bsky.social looked at how climate change poses risks to US drinking water utilities & if their bonds disclose these risks to investors. Utilities serving 67 million people have high risks, but 36% of their bonds don't mention climate: www.nature.com/articles/s43...

Asking Github copilot in agent mode to do a root cause analysis with search has been very helpful to get it unstuck on Rust projects

This is not true; I beg people read the full paper and especially the study design.

The conclusions mirror my (and many other practitioners') conclusions: if you use AI critically and engage both with the question and the answer, it has a net positive impact on both learning and productivity

Stateful Agents: It's About The State, Not The LLM

New post about why Moltbook freaks me out. tl;dr it's that you can't conflate a stateful agent with the LLM. Too much changes when you add state.

this one should be easier to read, more diagrams and easy to navigate text

timkellogg.me/blog/2026/01...

The breakthrough came when I stopped thinking about Atlas as a tool to configure, but as a system to design. Not “what instructions should I give?” but “what conditions produce good thinking?”

Post by @village11.bsky.social about how to build an autonomous agent

imo this snip right here is the distilled form of the difference between Clawedbot vs the agents @cameron.stream would have us build

www.appliedaiformops.com/p/what-build...

One Human + One Agent = One Browser From Scratch embedding-shapes was so infuriated by the hype around Cursor's FastRender browser project - thousands of parallel agents producing ~1.6 million lines of Rust - that they were inspired to take …

The latest entrant in the coding-agent-constructed web browsers is here, this one by @emsh.cat, and it's REALLY impressive - 3 days of development, 20,000 lines of Rust, no Cargo dependencies and it renders HTML+CSS extremely well
simonwillison.net/2026/Jan/27/...

yeah.. even the AI scene over there feels always unnecessarily angry at something or other, real or imagined

Been laughing at Claude for 5 mins

Been working with agentic document processing recently. Finding a reasonably sized powerful model is crucial and is something that doesn't tokenize. Doing it the old way - asking people on Discord.

ProPublica Publishes Unreleased Data on the Origins of Generic Prescription Drugs The dataset powers our Rx Inspector tool, which lets you find the factories that made your generic drugs and their FDA inspection track records.

ProPublica’s Rx Inspector allows you to find the factories where generic prescription drugs are made.

You can now see the previously unreleased data that powers our groundbreaking tool.