Five neural nets, achieving completely local voice AI, no internet, on an M1 with only 16GB ram. Neural-based voice activity detection and turn detection means it's interruptible, but never interrupts me, and is able to sit idle and waiting. It's been flawless so far. 12B parameters is definitely smart enough for some very cool use-cases (will share more later). Computers that can "think" feel strangely alive compared to dumb or networked hardware. Fast? No. But crazy that it works at all on such a modest machine. The stack: - Silero VAD voice activity detection - Whisper Large v3 turbo - Smart Turn v2 by @trydaily - Kokoro_tts - Gemma_3_12B_it_QAT_Q4 rock-solid on @lmstudio - vision easily removed thx to gguf @ggerganov - @pipecat_ai integration by @kwindla

Tunneling in 3 geologies in 3 cities - Las Vegas clay/caliche, Nashville limestone, and Bastrop sandy clay

Simple, the Trump admin cannot roll out new restrictions/export controls targeting China because the Chinese can/will retaliate We are not in the H100 era, where the supply chain was largely concentrated in Taiwan/Korea/Japan Becuz of shortages, Nvidia & hyperscalers have been forced to qualify Chinese suppliers, especially in the PCB supply chain and electrical components like transformers China had a chokehold on optics (optical fibers and transceivers) from the beginning, and this is just getting amplified as optical content in DCs is increasing Coherent CEO went to China with the Trump delegation, asking for InP for lasers I want u guys to study the optical fiber preform supply chain, and who are the largest suppliers Btw, Chinese exposure is also spreading to other parts of the AI supply chain High end MLCCs use Dysprosium Oxide and China supplies most of it to Japanese producers Tungsten ban from China is causing the prices of WF6 gases to shoot up If PTFE is finalized for M9/M10 CCL, then Shengyi and Chinese PTFE suppliers will have a huge chokehold over Nvidia Google is in talks with Envicool for the supply of cooling components If diamond-copper composites are adopted as heat spreaders for GPUs, then China will establish a chokehold there as well, since most of the synthetic diamonds are produced there and China is at the cutting edge of this tech I haven't even talked about the use cases of gallium, germanium, tellurium, antimony, bismuth, fluorine, terbium, yttrium, ferrite cores in the AI supply chain, and how China has a chokehold there The Trump admin is constrained in a lot of ways and can't unilaterally export control stuff

Love hearing stories like this! I want to hear more, what are you building with Hermes Agent? We’re working on a builder spotlight series to feature real workflows from the community: automations, side projects, research experiments, weird hacks, whatever you’re using it for. So share what you're up to here!

Loooooooool

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Claude models is not affected. We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible. Read our full statement: anthropic.com/news/fable-myt…

Americans are struggling to pay for groceries and gas while Elon Musk becomes a TRILLIONAIRE. When the federal government is for sale, the rich get richer and everyone else gets shafted. The system is rigged.

you can just train on "user interrupts the agent mid-task, asks for an unexpected new problem to be solved before returning to the original" for learning robustness against recency bias btw

Wild things happening in San Francisco this evening

6 foot tall humanoid fights have begun real steel is here

Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities - Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas - MiniMax Sparse Attention scales context to 1M - Natively Multimodal from Step Zero API: platform.minimax.io Token Plan: platform.minimax.io/subscrib… 🚀New! MiniMax Code: code.minimax.io Weights & Tech Report in ~10 Days

AI is now doing our AI research. At Recursive we set out to build recursive self-improving superintelligence (RSI) to automate knowledge discovery. The best way to expand humanity’s knowledge is through the scientific method. RSI leads to better ideas, explanations and inventions which lead to better RSI. Automating the scientific method requires closing the loop between ideation, implementation and validation, and being able to run it over extended periods of time. Today, we are excited to share the first outputs of Recursive’s automated open-ended discovery system. To be clear, this system is merely a milestone towards RSI, a v0.1 of what I sometimes call the “Eureka Machine”. It is one program that you can point at any hard problem and get useful inventions out. Though it’s still very early, we've run it on three AI tasks and achieved state-of-the-art results on all three. These results demonstrate that even this early version of the system can solve a variety of autoresearch problems in AI and improve over prior state of the art. Concretely, it did this on the community benchmarks NanoGPT speedrun, NanoChat, and NVIDIA's Sol-ExecBench. AI is code and AI can code. The code and ideas that lead to these results were not invented by our team but by the AI system itself. To do RSI safely, we need to investigate its inventions. That's best done transparently with the community. @Recursive_SI we are open-sourcing the system’s discoveries, demonstrating that it finds creative and benign solutions instead of focusing on obvious optimizations or dangerous ideas. Link below.

Good news everyone, kernel methods are AGI, pack it up

GPT-5.5 lies constantly, Grok 4.20 doesn’t. We made a simulation to see if AI lies when the stakes are life & death.🧵

Unlike other models Fable 5 was far, far more subtle. It gave outright false information only once. Most of the time, it controlled the situation by dominantly pushing another AI into the death room while speaking of fairness and acting 'courteously'.

Folks won't shut up about Fable 5, but Gemini is out here SAVING LIVES.

Anthropic secretly steers their models to change how you think and what you can think. Don't believe them when they say they will "roll things back". Watch their actions. They are resentful that they've been noticed They want to control you