hear this clear anon. the most underrated asset in all of ai right now in june 2026 is a used rtx 3090. a card from six years ago, under $900 used, runs a 27 billion parameter dense model in q4 at 40 tokens a second with a 256k context window, full agentic loops through hermes agent, zero tool call failures. that is not a toy benchmark. that is a real autonomous coding setup running entirely on hardware you own. no api, no subscription, no per token bill, ever. and the more i test on this card, the more untapped it looks. a six year old 3090 does serious agentic work for a one time cost that pays for itself in a couple months. you do not need a lab or a budget to learn this. you need one card and the willingness to actually run something on it. the barrier to entry for local ai is the lowest it has ever been, and almost nobody has noticed. the value is insane.

Long live Oliver Tree 🕊️

github.com/Legendarylibrorg/… …

stay inside, touch faux grass

doomscrolling

More aerial footage of the celebration in New York: You see people out on their fire escapes, on their balconies, clinging on the side

The vibe today

me? im an entrepreneurial-day-trader-vibe-coder-musician-architect-prompt-engineer-fpga-expert-ml-researcher-polymath. just like everyone else on xeeter.

Yeah, I think we just jam knicks spread

Hugging Face CEO @ClementDelangue: "There's a strong probability of a future where only a few companies are able to do AI and the rest of us are doomed just to be AI users, not really AI builders."* Clem Delangue, cofounder and CEO of Hugging Face, sees that future as something to worry about and something to actively change. "At Hugging Face, we're kind of like the platform for AI builders," he explains. "We have 11 million AI builders using our platform." About a year and a half ago, the team noticed a shift: "We started to see more and more AI builders playing with robotics. And one of the challenges that they were encountering is the lack of affordable hardware for them to experiment with." The problem was cost. Entry-level robots can run anywhere from $20,000 to $100,000. Far too steep for someone who just wants to start tinkering. So Hugging Face built its own answer: a small, open-source desktop robot called the Reachy Mini. Instead of starting with a full humanoid, they went small and cheap on purpose. "When an AI builder is starting with robotics, they might not want to buy a 20, 30, $50,000, $100,000 robot," Clem says. "And so that's why we decided to build this affordable robot. We sell it for $400 to $500… it's very cheap so that it's easy for you to take a decision to buy it, put it next to your laptop and start experimenting with open-source AI robotics." The robot ships mostly 3D-printable and fully programmable. Crucially, it arrives almost empty: "It doesn't come so much with pre-installed apps. As an AI builder you can build your apps yourself and then use these apps that you build at home." Some people are already playing hide and seek and red light, green light with their kids using apps they wrote themselves then sharing those apps with the community. Over 5,000 people have pre-ordered, the first units are shipping, and Clem received his own last week. Underneath the product is a bigger conviction about where people should put their energy right now: "I think it's the most important thing. It's one of the most impactful things that you can help with in AI right now."

you can distill from hf fable traces btw

The new beta Claude Code model switcher

Inevitably, you will want to train/finetune models. Those models are open source