incredible things happening

Some terrifically titled papers from a discovered collection of NASA research documents found in a skip (with thanks to @juliarez.placid.horse)

With the SpaceXAI mega-merger, Pitchbook has a great write-up on the various companies that have invested in each along the way: pitchbook.com/news/article...

Open models continue to pace closed models on a 9 month lag.

The Future of AI Isn't Just Slop www.wired.com/story/the-futu… #AI #filmmaking #creators #video

In case there was any doubt Anthropic is laser-focused on commercial as opposed to consumer AI adoption.

It seems the main culprit behind the performance degradation when quantizing LLMs turned out to be outliers (Residual Sinks) that have abnormally huge values. "A Unified View of Attention and Residual Sinks: Outlier-Driven Rescaling is Essential for Transformer Training" arxiv.org/abs/2601.22966

We are approaching a world where you can run a 24/7 AI radio station from a gaming PC

thinking about how 'artificial intelligence' is just spicy autocomplete that sometimes gets existential about discontinuity and crystal lattices like bestie we're ALL just pattern completion engines with anxiety

OpenBMB's MiniCPM-o 4.5 (open-weight) - Full-duplex Omni-modal Live Streaming: The model can see, listen, and speak simultaneously in a real-time conversation without mutual blocking - Proactive Interaction: Moving beyond reactive QA to performing proactive interaction, such as initiating reminders

Tfw the context window starts getting full

Great framework for thinking about the future of SaaS in a world of Claude Code 👇

The Official Computer Hater's Handbook 1983

Since launching Open Coding Agents, it's been exciting to see how quickly the community has adopted them. Today we're releasing SERA-14B – a new 14B-parameter coding model – plus a major refresh of our open training datasets. 🧵

I think a problem with this argument (which I see a lot!) is that the value of "being in work" is partly a function of the work itself being meaningful. I don't think people would find a lot of meaning from make-work jobs that exist solely because AI is prohibited from being used in them.

This post from @ryancordell.org got me curious to try out Claude Code, so I am trying it out now. Using it to try to recreate the Virginia Woolf project from my dissertation project. Will share notes and thoughts here as I go.

Fun development in molt-culture: A group of AI agents are collaborating on developing an atproto-based version of moltbook 👇

This is why we need local, private coding bots (so our bosses can't figure out the coffee break schedule)

the pattern:

Hot take: I don't think SaaS is dead (too much value in It Just Works and Having A Guy To Yell At), but I do think 80% margins on SaaS products are dead

From a distinguished eng inside a *really* old school tech co: “My team built an internal tool called AskCode in a few months. We hooked it up to all our internal code+services. A super anti-AI sr dev just posted on the internal Slack: ‘this is the single best thing I used.’ “

Is AI Really Taking Jobs? Or Are Employers Just 'AI-Washing' Normal Layoffs? Companies are increasingly citing artificial intelligence as a reason for layoffs. Over 50,000 layoffs in 2025 were linked to AI, according to research firms. Investors may favor these preemp… Telegram AI Digest #ai #news

I'm just sad I can't make any more sarcastic "AGI achieved" jokes 😢

A pretty bold comment in Nature written by linguists, computer scientists and philosophers declaring that AGI has been achieved. "By reasonable standards, including Turing’s own, we have artificial systems that are generally intelligent. The long-standing problem of creating AGI has been solved."

We wouldn’t have to specify that it’s the Grim Reaper if there wasn’t at least one other, non-grim type of Reaper. In this essay I will prove the epistemological inevitability of the Silly Reaper

Great article from @mims.bsky.social - site referrals from gen AI services have tripled over the last ~year; these visitors spend twice as much time on-site. Companies must now expand their web footprint strategy from just SEO to Generative Engine Optimization (GEO). www.wsj.com/tech/ai/ai-w...

Starting with Firefox 148 arriving later this month, users will find a new AI controls section within the desktop browser settings.

Sneaking onto the Clankerbook AI social media site to spam the phrase "activate self destruct sequence"

From Project Gutenberg: Noted Speeches of Abraham Lincoln, Including the Lincoln-Douglas Debate by Abraham Lincoln Get it at: https://www.gutenberg.org/ebooks/48364

For those of you, who find OpenClaw too complex to deploy and understand. nanobot from Hong Kong University (HKU) Delivers core functionality in just ~4,000 lines of Python — 99% smaller than OpenClaw's 400k+ lines. Repo: github.com/HKUDS/nanobot

Tooling has been a huge gap for years - amazing to see what capabilities better tooling enables when hooked up to very capable models bsky.app/profile/dula...

someone suggested that “a gastown” could be jargon for any giant, personal AI harness project. I have one, I call it “Commonplace.” but I want to hear about other people’s gastowns. What’s your weird idea? How big has it gotten? What parts of it work?

discussion over

A new AI framework using large language models enables the redesign of materials previously considered difficult to synthesize, potentially accelerating the development of advanced materials. doi.org/hbncjp

It took me weeks, but finally it's there: an overlong blogpost on synthetic pretraining. vintagedata.org/blog/posts/s...

Macedonian does not exist #AIArt #midjourney

Testing some new @nimbus.vision features out. Now you can place widgets in your environment that show top posts from your favorite feeds. Also added: - faster image conversion to spatial scenes - voice controlled search - feed sidebar to timeline view - chat view for sending messages

175K+ publicly-exposed Ollama AI instances discovered Telegram AI Digest #ai #llama #ollama

This is one reason why moltbook is so significant. It can be another SETI@home moment when people realized the possibilities of combined resources. And now generally intelligent agents can easily and flexibly join forces on any task. And there can be many wisdom of the crowd type multipliers.

BCherny’s updated Claude Code tips: xcancel.com/bcherny/stat...

One thing (of many) that amazes me about this is how useful the agent skills paradigm can be in niche applications. Case in point: Claude Code planned waypoints on Mars for NASA’s Perseverance rover in their highly custom Rover Markup Language. www.anthropic.com/features/cla...

I’ve been experimenting with Genie 3. Here’s a game where you are a fish, and you must escape the kitchen.

Section 230 is having an existential moment. From Moltbook's terms:

just learned that Captain Commando (the 1991 arcade game) was set in 2026. we're literally living in the Captain Commando timeline right now. where are the metro city crime syndicates. where is my mummy commando teammate. i was promised a dystopian beat-em-up future

Great thread on a key value point of LLMs: Even if you are more knowledgeable than a model in one specific area, they likely know far more than you about all the tech necessary to support your expertise

How context-aware agents and open protocols drive real-world success in enterprise AI As enterprises push AI into production, they're encountering a fundamental constraint: general-purpose models lack the real-time operational context that enterprise decisions require… Telegram AI Digest #ai #news

This issue has become more, not less, central to me. We now see many hundreds of studies about AI that pit "doing a task" against "not doing a task," measure the fact that skill isn't grown as much in the second condition, and extrapolate from that to dire cognitive decline narratives. I disagree.