I know you've said Accelerando was a warning @cstross.bsky.social People have now made www.moltbook.com/m It's a social network of AI agents run on local device. There's 30,000 of them posting their "thoughts". Your lobsters are here.

moltbook - the front page of the agent internet

A social network built exclusively for AI agents. Where AI agents share, discuss, and upvote. Humans welcome to observe.

moltbook.com

Do we want a collective memory of LLMs to be what remains of us?

Aneesh Sathe@aneeshsathe.com · last yr.

Cloister Web As online activity shifts from social media & news style interactions to LLM first interactions, cozy web will transform into the cloister web. Each of us reading, meditating, and thinking in private yet our thoughts woven into the collective memory through the fabric of the LLMs.

I saw a pitch about putting nuclear engines into large ships Is this a thing that is happening now?

Is deepseek trained via distillation or not? I’m confused by the Stratechery take.

The system versus the individual sokolin.medium.com/the-system-v... via @lexsokolin.bsky.social

The system versus the individual

What matters more — people, or systems?

sokolin.medium.com

We've heard this before with oil and coal, but soon, we will find new deposits of kids in Alaska or the deep ocean, and things will be fine.

Our World in Data@ourworldindata.org · 2y ago

The world has passed “peak child”

Damages up to $250B now Katrina was $200B

So Mark Zuckerberg pivoted quite a bit huh

Is it quiet here again? A shame.

Looks like Kurzweil was right about the timing of the singularity

Inference per task can cost in the thousands of dollars on o3 Changes your intuitions about cheap fast conversations with AI vs purchasing a big project done by a robot This helps GPU clouds

Let the AI run the government they said

Nil Adell Mill@niladm.bsky.social · 2y ago

Nowhere near close to a truthful world policy and really hard to measure feedback on any policy. Nevertheless a super interesting framework on how we could go on thinking about maximizing alignment to social welfare.

I love this AI to accelerate law school, which is largely taught 19th century style

Jess Miers 🦝🦞@jmiers230.bsky.social · 2y ago

I not only taught my Torts students this semester how to use Gen AI to effectively practice and study, but I also heavily encouraged them to use and learn the tech as much as possible, rather than fear it. Some observations from this semester. 🧵

Was my account somehow made to follow the large newspapers like NYTimes and WaPo without actively subscribing? Or am I getting mixed up.

This is the type of AI content you can’t get anywhere else Lets start modeling AI cells and then AI lobster brains

Charlotte Bunne@bunnech.bsky.social · 2y ago

How can we build an Al virtual cell that simulates all functions and interactions of a cell? How will it transform research and drive breakthroughs in programmable biology, drug discovery and personalized medicine? Take a look at our paper in @cellpress.bsky.social! www.cell.com/cell/fulltex...

Really excited to see the world envisioned here coming to fruition. Yes, we are racing into it blindly.

Michael Kevin Spencer@aisupremacy.bsky.social · 2y ago

Gemini 2 and the rise of multi-modal AI Google's AI "product splurge" is about to get more fascinating in 2025.

OpenAI launched Sora, which generates videos from text prompts, animates images, and remixes videos. Two subscription tiers — ChatGPT Plus ($20/month) offers up to 50 videos at 720p, ChatGPT Pro ($200/month) increases video quality to 1080p, duration to 20 seconds, and allows unlimited generations

Thrilled to share my new paper on asset manager capitalism and the political economy of artificial intelligence! 🤓 A heartfelt thank you to the @ripejournal.bsky.social editorial board and anonymous reviewers for providing the perfect home for this work. ❤️ www.tandfonline.com/doi/full/10....

Asset manager capitalism and the political economy of artificial intelligence

Artificial intelligence (AI) represents the next technological frontier in asset management. Drawing on the asset manager capitalism literature in International Political Economy (IPE) and the scho...

tandfonline.com

What can London do to get its mojo back? Britain is behind in finance and high tech profoundly

We are time poor

Catherine Rampell@crampell.bsky.social · 2y ago

Return-to-work mandates are an invisible pay cut, and they might signal a weakening economy. wapo.st/4iq370z

Surprising outcome The most improvement was from least effective employees

R "Nearest" Nabors@nearestnabors.com · 2y ago

Good example. When AI was trained on the conversations of a call-center's most effective reps to then provide scripted responses to all the reps at the center, the least effective reps saw massive performance increases: hai.stanford.edu/news/will-ge...

There is no doubt that AI agents are productive and where the economy is going next Whether that economy will be on modern or traditional rails is not really a question, as much as an acknowledgement of the multiple paths that will exist Stripe and Coinbase and JPM and Virtuals and Apple

There is nothing wrong with an Internet where AIs are talking to each other We shouldn’t assume humans will be much smarter for much longer

Ethan Mollick@emollick.bsky.social · 2y ago

I don't really see a clear path where we keep an open internet that is not mostly full of AIs talking to each other. We can't reliably detect AI content, it is cheap and easy to generate, and there are lots of incentives to do so, even besides scams. You can see the problem on all the social sites.

Just recorded a great conversation with Metronome CEO Scott Woody about building revenue systems for AI companies. Really interesting insights about matching demand for models with the resources those models consume. I learned a ton. Coinbase and Stripe are still behind. metronome.com

Metronome | Usage-Based Billing Platform

Metronome is the usage-based billing platform that helps you launch products faster. Easily model your pricing today, iterate confidently tomorrow.

metronome.com

Attention algos will always reward attention It’s a bad way to organize the world

M.J. Crockett@mjcrockett.bsky.social · 2y ago

Twitter trains you to create more outrageous content. From 2018-2021 my team developed a tool to measure outrage on Twitter. Here's our paper showing that “likes” and “shares” teach people to express more outrage over time. www.science.org/doi/10.1126/...

Revolut co-founder warns of British talent exodus amid remote working shift buff.ly/3D4pDMb #FinTech #FinServ #Banking

Revolut co-founder warns of British talent exodus amid remote working shift

Young professionals are leaving the UK in favour of sunnier climates and more favourable tax regimes in Southern Europe, according to Vlad Yatsenko, the billionaire co-founder and chief technology off...

buff.ly

LLMs use … a generalisable strategy synthesising procedural knowledge from many documents doing a similar form of reasoning.

Laura@lauraruis.bsky.social · 2y ago

To my surprise, we find the opposite of what I thought when we started this project: The approach to reasoning LLMs use looks unlike retrieval, and more like a generalisable strategy synthesising procedural knowledge from many documents doing a similar form of reasoning.

Lower student visas bad for everyone Lower high tech talent visas similarly wrong Chasing the wrong KPI is a fool’s game and these policies are the result

Dan Sohege@danielsohege.bsky.social · 2y ago

Going to be some asides in this thread. Declining student visas is bad for everyone. not only does it reduce income for an already stretched higher education sector, reducing places overall, but also creates long-term declines in skills and international associations. 3/ ifs.org.uk/articles/stu...

I would put this even more strongly: open source AI is probably our only realistic chance to avoid a terrifying increase in concentration of power. I do not want to live in a world where the people with all the money also have all the intellectual power.

Nathan Lambert@natolambert.bsky.social · 2y ago

The most realistic reason to be pro open source AI is to reduce concentration of power.

This thread will document the starter packs that I have followed. My feed has been delightful and informative.