Crazy what it takes to get a decent phone picture of the Moon

Screenshot of a table. There is too much information in the table to fit within the character limit for this alt text. You can find the table in this reddit post where your screen reader will work nicely with it: https://www.reddit.com/r/murderbot/comments/1sp5s88/a_straightforward_guide_to_all_the_murderbot/

My friend is reading The Murderbot Diaries by @marthawells.com, which inspired me to finally make this guide to the series.

I'm pretty proud of this table. I tried really hard to make it easy to understand, since most sources for this information are pretty confusing.

reBot: $1k Budget to Build Open-Source 6+1DoF Robotics Arm for Physical AI YouTube video by Seeed Studio

Seeed Studio’s reBot Arm B601 looks *so* good. Actually nice actuators (no more servo jitter!), and robust mechanical components. The price: ~$1k. Sensing is an evolving story, but this looks like a great basis for working with VLAs even at the home lab scale.
youtube.com/watch?v=ONbp...

A Steerable Model with Emergent Capabilities A steerable robotic foundation model that exhibits a step-change in generalization.

I like how Physical Intelligence demonstrates what their model can do with ~modest hardware. Even if the behaviors aren't ready to roll out into a production setting, it's important to see what can be done with things like parallel grippers rather than expensive hands. www.pi.website/blog/pi07

System Collapse: The Murderbot Diaries Am I making it worse? I think I'm making it worse.

#DEAL ALERT! Calling all one-clickers because SYSTEM COLLAPSE @marthawells.com is on #sale on #Kindle US for today only!
amzn.to/4tEXJvc
#SystemCollapse #TheMurderbotDiaries #ad #amazoncom

1/n 🧵 Introducing Gaussian Wrapping — a principled framework for extracting high-quality meshes from 3DGS! 🚲
We recover thin structures, like bicycle spokes, where all prior methods fail.

Follow the thread for a brief overview and links!

Love that the debate about the value of GenAI doing things humans enjoy doing (eg art) is expanding to consider the value add of GenAI exploiting security vulnerabilities.

Boy, I had a Claude session end with Claude complaining that it had worked quite hard on fixing a bug and thinks we should give up followed by a session where it failed to use consecutive MCP tools by misnaming required arguments. That's... not great.

It's not clear to me that the workflow of moving between spec, proof, deploy is really setup for iterative refinement the way I'd want here, but it's got a lot of good pieces!

GitHub - jasisz/aver: Aver is a programming language for auditable AI-written code: verify in source, deploy with Rust, prove with Lean/Dafny Aver is a programming language for auditable AI-written code: verify in source, deploy with Rust, prove with Lean/Dafny - jasisz/aver

Expecting to see more like this, but good to really think about it: “Aver is a programming language for auditable AI-written code: verify in source, deploy with Rust, prove with Lean/Dafny”

github.com/jasisz/aver

Children at an Edinburgh hospital painted pebbles for the penguins at Edinburgh Zoo and will be able to watch them choose and present them to their mates via livestream.

#SundayMorning

Interview: Inside Serve Robotics -- How Autonomous Delivery Robots Navigate Cities YouTube video by Learn the Technology with Brandon Krakowsky

Talked about delivery #robotics on a technology interview series. I hope it’s a decent overview of how we approach the business! youtu.be/cR7migPW9Fo?...

Happy St. Patrick's Day, New York.

paris jackson - zombies in love YouTube video by ParisJacksonVEVO

I’ve worked in #robotics for a long time. This is a new one for me. www.youtube.com/watch?si=cyY...

Queen Demon (The Rising World Book 2) Amazon.com: Queen Demon (The Rising World Book 2) eBook : Wells, Martha: Kindle Store

There's an Amazon Kindle sale on Queen Demon (the sequel to Witch King, 2nd book in the Rising World series) today for 2.99, US and Canada www.amazon.com/Queen-Demon-... #booksky

An almost aggressively jaded viewpoint comes through the text of that article.

This is a very niche observation, but regarding Apple announcing a bunch of new hardware... thinking we'll see a new Apple TV with a faster processor coinciding with the release of Oceanhorn 3 on the 5th.

I wonder the extent to which Anthropic is able to ask Claude for help diagnosing and resolving the current Claude outage.

do you code goofy or regular?

I really hope you enjoy it!

I did enjoy Generosity. Goldbug shook me both with the book on its own merits and seeing that a person could write that book. It’s a ridiculous display by an author.

A big issue with this is when the agent doesn’t implement what you discussed. You then end up in this situation where a later development relies on something you thought was there, and it turns out it was totally missed. There’s no responsibility on the agent’s side, it’s all on you.

Short musings on "cognitive debt" - I'm seeing this in my own work, where excessive unreviewed AI-generated code leads me to lose a firm mental model of what I've built, which then makes it harder to confidently make future decisions simonwillison.net/2026/Feb/15/...

Goldbug, though it was my first so perhaps that’s some bias, but it just floored me. Really enjoyed Orfeo and The Overstory, too.

I have a funny relationship with his writing in that I find it almost too powerful for comfort. I’ve read 5 or 6 or so of his books but not all, and the ones I haven’t read I’m not sure I want to!

Meme from the show Arrested Development with Tobias Funke saying, "Responsible proponents of LLM coding agents suggest that you have to approach their use with caution. Steer them with structure and organization." His wife, Lindsay, asks, "Well did it work for those people?" Tobias responds, "No. It never does. I mean, these people somehow delude themselves into thinkint it might. But... but it might work for us."

As much as I think LLM coding assistants can be great aids, it is reaalllly hard to have it not come across as a Tobias take.

Scaling long-running autonomous coding Wilson Lin at Cursor has been doing some experiments to see how far you can push a large fleet of "autonomous" coding agents: This post describes what we've learned from …

Having compiled and run the web browser that Cursor built in a couple of weeks using mostly a giant fleet of coding agents I'm actually very impressed by it - there are rendering glitches but the renders it produces are surprisingly usable for a few-week-old project simonwillison.net/2026/Jan/19/...

Mario Kart World seems to have a weirdly uncomfortable relationship with its platformer side. Why not make split screen Free Roam less hidden, and why not embrace the THPS influences and make charge jumps easier to activate?

Hard to avoid. You can buy extra usage if it really is a one time thing, but a couple days of not wanting your schedule dictated by that session timeout is remarkably well calibrated to push you to Max.

It may be driving us towards programs that are more amenable to machine verifiable proofs. That’d be a nice outcome!