Your curated collection of saved posts and media
yeah this is where it starts getting fun. this morningโs journey log: getting hermes running with mixture of agents across 2 dgx sparks. one spark gets qwen. one spark gets nemotron. then hermes sits in the middle and uses both instead of me depending on one model to handle everything. thatโs the part iโm testing this morning two boxes on my desk, a bunch of local models, and me trying to see how far i can push this thing. local ai is starting to feel less like โrun one modelโ and more like building your own little team of brains.
ใๆก็จๆ ๅ ฑใใใญใฐใฉใ ใใใผใธใฃใผ(RSI Lab)ใฎใใธใทใงใณใใชใผใใณใใพใใ๐ https://t.co/57Eyb23kws RSI Labใฎ็ ็ฉถๆดปๅใๆฏใใใใญใฐใฉใ ใใใผใธใฃใผใๅ้ใใพใใใใใใฏใฉในใฎ็ ็ฉถ่ ใปใจใณใธใใขใ็ ็ฉถใซๅฐๅฟตใงใใ็ฐๅขใใคใใๅฝนๅฒใงใใ ใใฎใใใชๅฝนๅฒใๆ ใฃใฆใใใ ใใพใใ ใปไบ็ฎ็ฎก็ใปในใฑใธใฅใผใซ็ฎก็ใๅซใ็ ็ฉถใชใใฌใผใทใงใณๅ จ่ฌ ใปใชใตใผใใฃใผใจๆ่ก็ใชไผ่ฉฑใใใชใใใ่จ็ปใจๅฎๆ ใฎใฎใฃใใ่ชฟๆด ใป็คพๅคใใผใใใผใจใฎ็ชๅฃใจใใฆใฎใณใใฅใใฑใผใทใงใณ ไบ็ฎ็ฎก็ใใใญใธใงใฏใใใใธใกใณใใๅฏพๅคๆ่กใชใฉใฎๅฎๅ็ต้จใใใใใใธใในใฌใใซไปฅไธใฎ่ฑ่ชๅใใๆใกใฎๆนใๆญ่ฟใใพใใ ็ ็ฉถใๆฏใใ็ซๅ ดใใใAIใฎๆฌกใฎใใฉใใคใ ใฅใใใซ้ขใใใใๆนใใใฒใๅฟๅใใ ใใ๐
Iโm looking to hire a Program Manager to help manage Sakana AIโs fast growing Recursive Self-Improvement (RSI) Lab ๐ RSI Lab (English): https://t.co/Sz46xHIfZi RSI Lab (ๆฅๆฌ่ช): https://t.co/gCBtwg5imq Job Description: https://t.co/DoWrN7jSl9
ใๆก็จๆ ๅ ฑใใใญใฐใฉใ ใใใผใธใฃใผ(RSI Lab)ใฎใใธใทใงใณใใชใผใใณใใพใใ๐ https://t.co/57Eyb23kws RSI Labใฎ็ ็ฉถๆดปๅใๆฏใใใใญใฐใฉใ ใใใผใธใฃใผใๅ้ใใพใใใใใใฏใฉในใฎ็ ็ฉถ่ ใปใจใณใธใใขใ็ ็ฉถใซๅฐๅฟตใงใใ็ฐๅขใใคใใๅฝนๅฒใงใใ ใใฎใใใชๅฝนๅฒใๆ ใฃใฆใใใ ใใพใใ ใปไบ็ฎ็ฎก็ใปในใฑใธใฅใผใซ็ฎก็ใๅซใ็ ็ฉถใชใใฌใผใทใงใณๅ จ่ฌ ใปใชใตใผใใฃใผใจๆ่ก็ใชไผ่ฉฑใใใชใใใ่จ็ปใจๅฎๆ ใฎใฎใฃใใ่ชฟๆด ใป็คพๅคใใผใใใผใจใฎ็ชๅฃใจใใฆใฎใณใใฅใใฑใผใทใงใณ ไบ็ฎ็ฎก็ใใใญใธใงใฏใใใใธใกใณใใๅฏพๅคๆ่กใชใฉใฎๅฎๅ็ต้จใใ
@repligate Here is where it sounded like after a couple compress cycles after I asked it to build the "last person shooter" (from a viral post mentioning the idea) themed around Durkheim and Weber. Advanced Claudish. https://t.co/oQv2OfHoSO
*Cracks knuckles* https://t.co/EP9lTnbt5W
@tshebetsot92 @Hesamation Yeah I would say so, otherwise it would be 3x as long. But I have a primer here that might help: https://t.co/OpJnPkrGK9
One thing I mentioned only in passing in my Fable post is that, for long running tasks, Fable starts to develop its own dialect as its many agents and tasks reinforce themselves and make Claudish language ever more Claudish. You need to ask it to report out in plain English. https://t.co/vJ5111DE2E
Fable, one prompt: "build an elaborate game that makes it feel like I'm a brilliant chess player without knowing anything at all about chess. It should make me feel like a grand master. Feel free to go as meta as you want but the more chess-y the better." https://t.co/XYVsgQpzPI https://t.co/ofRLmcf57j
@PatrickMoorhead https://t.co/uaFKUUl89h
higgsfield is LYING TO YOU it's true. they're an AI wrapper that charges you tons of credits for using the same API everyone else uses, and way more expensive what's the point in paying for monthly credits when you can just run Replicate and have the same results? i get it, they actually did a great job at wrapping things together and marketing it with tons of ads, but don't be fooled stop wasting money and add your github to Replicate and pay per generation instead (way way cheaper!) btw, i'm not sponsored by Replicate but i felt the urge to tell you, even more if you're trying to increase your MRR higgscam will steal all your money and you won't even notice i post about what works and what doesn't, and how to actually save money on marketing bc most of us don't have that much cash $$$ what do you think? did you run out of credits? will keep you updated
Meta's AI infrastructure may be about to do more than power its own products. By selling excess computing capacity, it's turning a massive investment into a potential new business. That could become a model others choose to follow. https://t.co/b97xBCtWpg
Very proud to have spoken at @aiDotEngineer! Talked about automating my job at @huggingface with agents ๐ฅท Involves: > Claude Agents SDK > GLM-5.2 via Inference Providers > @langfuse for tracing > @modal for deployment Will be available on @YouTube later https://t.co/cnGN3hWrNO
GLM 5.2 DSpark preview is here! โจ https://t.co/DQOMYEiY1o This is the first DSpark speculator for a non-DeepSeek frontier model, trained with Speculators and running on vLLM nightly for ~1.5ร faster decode for GLM-5.2-FP8 on 4รB300. Stronger checkpoints to come!
this means GLM 5.2 DSpark on the way btw
GLM 5.2 DSpark preview is here! โจ https://t.co/DQOMYEiY1o This is the first DSpark speculator for a non-DeepSeek frontier model, trained with Speculators and running on vLLM nightly for ~1.5ร faster decode for GLM-5.2-FP8 on 4รB300. Stronger checkpoints to come!
Teams recently completed build and acceptance testing of our 1,000th Merlin 1D engine for Falconโs first stage! With Falconโs reusability, recovering these engines has enabled continued reliability enhancements, making Merlin one of the most reliable rocket engines ever manufactured
New article available to subscribers. https://t.co/4p9Y5lJWde
TESLAโS NEXT-GEN OPTIMUS V3 HAND IS GETTING CLOSE TO HUMAN-LEVEL Tesla engineers just shared new details on the Optimus V3 hand โ and the progress is striking. Theyโre now saying that as they move into Gen-3 and mass production, the hand is getting very close to human functionality and form factor. One engineer described it clearly: โIt wonโt even look like a robot. It will look like a human in a superhero suit. It will be something revolutionary.โ This level of dexterity and human-like design is a critical milestone. The hand is one of the hardest parts of building a truly useful humanoid robot โ and Tesla is iterating extremely fast.
unprompted btw ๐ญ https://t.co/Nc3BoGHSaA
unprompted btw ๐ญ https://t.co/Nc3BoGHSaA
now available in OpenCode Zen https://t.co/aZkAnPbRN3
Introducing ZCode, the official development environment for GLM-5.2 - GLM Coding Plan subscribers: now 1.5x usage quota in ZCode - BYOK supported: works with your existing subscriptions and APIs - Available on macOS, Windows, and Linux Download now: https://t.co/Peepqv4XSx
๐จ TRUMPโS FINANCIAL DISCLOSURE JUST DROPPEDโฆAND ITโS WORSE THAN YOU THOUGHT. The U.S. Office of Government Ethics released Donald Trumpโs 927-page financial disclosure today. Hereโs what a sitting U.S. President made in 2025: ๐ฐ $635 MILLION from the TRUMP memecoin, while retail investors watched it crash from $74 to $1.68 ๐ฐ $594 MILLION from World Liberty Financial token and stablecoin sales, a crypto venture co-founded with his own sons ๐ฐ $65 MILLION from selling equity in that same company ๐ฐ $80+ MILLION in media settlements from ABC, CBS, Meta, and YouTube, paid to his own presidential library Total crypto haul: over $1.4 BILLION. In one year. While serving as President of the United States. His net worth has nearly TRIPLED, from $2.4 billion to $6.3 billion, since taking office. And while Trump pocketed $1.4 billion from crypto, the everyday Americans who bought his memecoin? They lost. 764,000 wallets ended up in the red. This is the most corrupt presidency in American history. Period.
Elon Musk gets mad when people point out that DOGE and its USAID cuts killed thousands, if not hundreds of thousands, of babies, children, mothers, and elders. https://t.co/UdvkmcUnF1
$2.2 BILLION for the Trump family. Higher costs for the American people. https://t.co/kvtw061jcw
President Trump reaped at least $2.2 billion from his familyโs holdings last year, a new filing shows. https://t.co/Rdp8KHJLxl
$2.2 BILLION for the Trump family. Higher costs for the American people. https://t.co/kvtw061jcw
Fugu is now available on OpenCode! โจ When our team was developing Fuguโs multi-agent orchestration, OpenCode was our tool of choice to verify our models. We share a core philosophy with the OpenCode team: the future of coding agents should be an open, collective ecosystem. https://t.co/rctKxD7jcE
Bro, @DarioAmodei I know running a company is ruff, but claiming you have superintelligence that could destroy economies and also claiming you canโt tell if this is a dangerous question is justโฆ well, lame - An early user of @claudeai and unapologetically positive AI person https://t.co/JmHyiQ5kZa
TTS Arena is live. ๐๏ธ A blind benchmark for text-to-speech, rebuilt from the ground up - listen to two anonymous models, pick the one that sounds more human, and watch the leaderboard move. Now with the latest models. Come vote โฌ๏ธ https://t.co/1hCCOES94Q
AI systems that are accessible or even better open-source are safer for a simple reason: more people can inspect them, test them, stress them, and report what breaks or harms to keep the builders accountable. "Sunlight is the best disinfectant"! Thatโs why weโre proud to help lead FLARE with a coalition of safety and cyber-security researchers from MIT, Stanford, Princeton, Harvard, Northeastern, Carnegie Mellon and more. The first release is a standardized way to report AI flaws across the ecosystem. One report can reach the right developers, safety organizations, and registries, so a flaw found once can be fixed everywhere faster with the right accountability: https://t.co/0V1tsdzjFq
โAgentic kernel optimization is the future of on-device inferenceโ @xenovacom used Fable 5 to write kernels that pushed Gemma 4 to a massive 255 tok/s on WebGPU with M4. He shared the demo, so you can try in your browser!! https://t.co/xPuh5OLGEt
HuggingChat inference on gemma-4-31B at 1x speed ๐คฏ https://t.co/j907DMS29A
HuggingChat inference on gemma-4-31B at 1x speed ๐คฏ https://t.co/j907DMS29A