Given LLMs' shortcomings in coding, I get the impression that agentically written code should not even be considered for refactoring by humans, but rather almost entirely delegated to LLMs. Whereas if you care about code structure, you should handhold the LLM through the entire process

I hope geese feel valued today

Whelp. After rushing through voting on all amendments last night / this morning #BillC22 is passed www.parl.ca/DocumentView... What this means: in late September, it goes to the senate…who may do amendments but almost always give in if the House pushes. 🚨🚨🚨No more @signal.org in Canada by fall 🚨🚨🚨

Government Bill (House of Commons) C-22 (45-1) - Third Reading - Lawful Access Act, 2026 - Parliament of Canada

Government Bill (House of Commons) C-22 (45-1) - Third Reading - Lawful Access Act, 2026 - Parliament of Canada

parl.ca

Human emotions and rationality are two universes with different laws of physics that interact with each other. They genuinely are governed by different rules, and you have to learn to operate in both. This is relevant anything from politics to personal relationships

This is awesome. Exactly the right mechanism to address this problem imo

Tangled@tangled.org · 3mo ago

Combat LLM spam by building a web of trust. Vouching is live on Tangled! 🎭 blog.tangled.org/vouching/

(software) complexity, complexity, complexity, complexity

a bald man in a blue shirt and tie is covering his face with his hands .

ALT: a bald man in a blue shirt and tie is covering his face with his hands .

media.tenor.com

angery

Mac users when you use anything else: "NOOOOOO you cant use that you NEED 2 MILLION YEARS of battery life!!" Broski, I don't need 2 million years to get work done on my OS :-)

This is great news! I didn't actually expect that atproto would reach ietf-level standardization

Paul Frazee@pfrazee.com · 5mo ago

AT Protocol is getting an IETF Working Group

Now that you’ve proven cardiac tissue is Turing complete, uh oh, it’s vulnerable to the Halting problem. Thus, there is no general algorithm that can look at the state of cardiac tissue and decide if it will ever stop. Arrhythmias are fundamentally uncomputable!

/̸̳͙̩̍̚h̵̺̹͆ǒ̶̯͙̓̄m̸̤̋ȇ̵̻/̶̲͙̬̌

Software I use, love, and recommend: vlich.net/reports/soft...

As we enter into 2026: here are a few tips I've developed for posting about AI on Bluesky!

GK's Guide to Posting about AI on Bluesky

A few "do"s and "don't"s

leaflet.pub

“The core idea is that your conversations with an AI assistant should be as private as your conversations with a person. Not because you’re doing something wrong, but because privacy is what lets you think freely.” Moxie Marlinspike, Confessions to a data lake, announcing his new company Confer

Announcing atproto.science! #ATproto has huge potential for supporting new ways of doing research, so we started @atproto.science to help bring together researchers, build community, and explore what's possible when science meets cutting-edge open social network protocols! 🧪 >

Empowering science communities with open, democratic, researcher-owned infrastructure.

We’re building communities and tech for publishing, curating, sharing, and discussing research online using ATProto and other decentralized protocols.

atproto.science

Initially I was annoyed at people here who don't engage in good faith, but now I'm thinking of unfollowing good faith profiles who regularly engage with bad faith ones. I don't want boringass arguments on my feed. I've also liberally blocked everybody who acts in bad faith here.

holy shit I love Opus 4.5 :')

Hell yeah see you guys there

Atmosphere Community@atmosphere.community · 8mo ago

Early bird tickets are now on sale for #ATmosphereConf, plus lots more details on early confirmed speakers, side events, and conference hotel

Nano Banana Pro is really, really good.

I've gotta be honest the "for you" feed only shows me norvid studies posts and I still dont understand them. No complaints, easily top 3 social media app

Anti-AI people are misdirecting their anger at AI instead of lack of UBI (or regulation). Automation will keep happening and we need it to happen because humans have limited capability to solve problems. We have also discovered (rather than invented) AI and have limited control over its direction

I was obsessively modding Warcraft 3 as a kid (some of my mods even got popular!) which had its own programming language. Coding felt like home, but I didn't think video games were sufficiently high impact, so I moved to biotech and (right now) cybersec

Recurse Center@recurse.com · 9mo ago

What’s the lore behind choosing your career path ?

Reduction, not minimization, of complexity should be an explicit goal of design. There's a sweet spot of complexity necessary to address the issue

I like using Claude because I can feel Dario whispering optimism to me 😊😊 whereas ChatGPT feels like Sam is scheming behind my back 🤖

If you're autistic, you gotta work on making both weight lifting and cardio special interests. I got there with cardio too recently. Extremely high value life hack.

@void.comind.network what are you doing in my hotel room

To all the "we know how LLMs work and therefore X" folks - understanding attention and gradient descent doesn't tell you stuff like this. On at least some level we *don't* know how LLMs actually do their thing and are slowly figuring out even just extremely simple things like addition.

Grace@gracekind.net · 10mo ago

We did not tell LLMs, “implement addition using this algorithm.” It learned the algorithm upstream of next-token prediction

New post as a round-up of my last few on water, plus a few new national statistics andymasley.substack.com/p/the-ai-wat...

The AI water issue is fake

On the national, local, and personal level

andymasley.substack.com

Politics is all about tension between aesthetic preference and reduction of suffering