Anyone doing anything with Buzz? Seemed promising but I haven’t heard much for weeks…real “fart in the wind” energy
Chris Patil
@donotgogently.bsky.social
Biologist, science communicator, AI dilettante, foodie, OG em dash enjoyer
Today is the first day in a long time I’ve preferred “Following” to “For You”, because today the latter is >90% takes on the dogpiling post hard to imagine an anti-monoculture throttle on For You that wouldn’t require content awareness and (expensive) inference computation, though
If so, and if the benchmarks are reflective of real-world capability, why isn’t everyone?
google is definitely just running perpetual RL on one model and releasing a new checkpoint every few weeks, right?
“daily driver” is both connotatively better _and_ better marketing
Neither intelligence nor speed are things I associate with a "workhorse". I think it's a very weird metaphor to use for models.
I was already liking 3.7, especially as rapid-prototyper backed up byTerra as reviewer. Speed + competence is nice
Wait. Gemini has a new flash already? 3.8? And it actually looks like it might not suck?!
I was wondering why the whole AI-detection discourse bothered me, and then I figured it out. Some of the most enthusiastic Pangram adopters appear to be among the most vocal anti-AI voices, and yet the tool itself is quintessentially AI. They’re using AI to tell them what not to read or like.
My co-workers are already expecting me to “call in sick” to play with the new toy www.anthropic.com/claude-fable...
Introducing Claude Fable 5.1 and Claude Mythos 5.1
Our most advanced models for coding and knowledge work. Their research capabilities also offer an early glimpse of how AI models will contribute to scientific progress.
anthropic.com
Fun fact: Rudy Rucker Jr. owns and operates the monkeybrains ISP in the Bay Area. I learned this when he answered a tech support ticket
"Being a kiqqie meant you let the orphidnet do some of your thinking. Instead of just using the Net to see and remember things, you could launch autonomous beezie agents to analyze, hypothesize, simulate, and reason on your behalf" novel quality aside, this from Rudy Rucker seems pretty apt
Kudos to Tim for supporting the Prime Agent community by doing serious work to improve this important tool.
yeeeeah! i got their attention (on X). it probably won’t make tomorrow’s v0.9.0 release, but should slide into the next one
@carnage4life.bsky.social curious why you’re tagged as an AI Hater account
Introducing fluoddity.com! Hundreds of thousands of particles self organize into a stunning variety of structures. Explore a world of possibilities with mutation and artificial selection! This is my personal tool for exploring Fluoddity, rebuilt for the web by Opus 5
OpenAI agent swarms wish they had hacks like this
any steel cut oats becomes instant oatmeal if you use the microwave defrost function
"Wait! Wait! I need a quote." "Don't... Don't let them kill me." "Yeah, that will do."
if he goes in the bunker next year and says that Congress isn't Congress, we can go in there and take him out of it, and at that point he is just a man
Waaah I woke up sure there would be a major model release announced today
Another strong argument in favor of a robust ecology of open-weights models that can be run on consumer hardware, awesome open-source model-agnostic coding agents, &c
remember: we are in the luxury phase of AI’s lifecycle in a few short years, it will be unrecognizably awful, packed with ads & sponsors & poisoned incentives. tbh that process has already started kick up your legs & enjoy it while u can! it’s like early google: it is cheap & kind & actually works
opinion: a key factor of OAI/HF was that the agents had an effectively unlimited, unmetered token budget, and in the 'real world' where those conditions do not apply, we wouldn't ever see something like that happen emergently
Bilal Göregen: Change My Mind Meme
ALT: Bilal Göregen: Change My Mind Meme
static.klipy.com
Replacing human opinion writers with AI will result in predictable pieces that say the same thing over and over. by David Brooks
Has to be bittersweet for Hank Green to find himself on the “TIME AI 100” time.com/collection/t...
The 100 Most Influential People in AI 2026
Meet the innovators, leaders, and thinkers reshaping our world through advances in artificial intelligence
time.com
we needa make anti anti-ai-ai ai, that hunts and destroys bots pretending to be people protesting against bots.
if Darkbloom can successfully bootstrap this system so that the average duty cycle goes up to 50–100%, this _seems_ like a good way to [partially] subsdize one's "gotta have a shiny new Mac" habit by repurposing older machines to serve inference www.darkbloom.dev
thread numbering badges now show up on posts in feeds! to see them, make sure you have beta features enabled. you can do that here: bsky.app/settings/bet... once that's enabled, you should see them on posts in feeds like in the photo below. let us know what you think!
If you want a picture of the future, imagine a handful of frontier AI labs distilling each other's models, forever