与「FLOP」相关的搜索结果

The spy hoax flopped. So they pulled a new rabbit out of the hat. Right when people finally started seeing the witch hunt against Jensen Huang for what it is— right when the sympathy turned— the Ivy League spin doctors ran the same play again. Same quiet weekend. Same meaningless signature stunt. Same goal: kill the heat and let Anthropic (@AnthropicAI) walk away clean. Please. I could sign Putin and Kim Jong Un onto that standwithjensen list myself. And you know what hurts? It worked. Again. I'm nobody. No clout. No machine. No PR firm. But I have a conscience. And my money goes where my conscience goes. Zero shares. Zero derivatives. Zero exposure. Not me. Not anything I own. Not a dime for $ANT.

Back in 2012, AlexNet was trained on two NVIDIA gaming GPUs. Jensen understodd what that meant before almost anyone else, and bet the company on it. A couple weeks ago he spent a good chunk of a meeting with us nerding out on our little $400 open-source microrobot. Why it runs hot, which chip would get more FLOPs, how many watts it would draw 🤓 His latest big bet has been open models, his very first post on X 👇 If open models had only one champion, this week would have been the end. They have millions of builders, including plenty inside NVIDIA who’ve been messaging us the same thing: keep building in the open. We will.

If I had $500k in crypto now, I would: 1. Go full-time crypto. Move somewhere cheap. One year to make it. 2. No BTC, no ETH, no SOL, no HYPE. 3. No memecoins. No directional leverage. 4. Allocate $250k to a delta-neutral strategy between @tradexyz and @metricxyz. -> Short equity perps trading at a premium on tradexyz, buy the matching tokenized equities on metricxyz. One strategy, three potential airdrops: HL S3 + tradexyz + metricxyz. 5. Allocate $50k to prediction markets on HIP-4. Market make tradfi outcomes. If options launch, be ready to arb them against a broker from day one. 6. Deposit and loop $200k of sUSDe on Aave. -> Deposit sUSDe, borrow USDC, swap into more sUSDe. Use the net yield and some of the principal to cover living expenses. Rotate into @acier, an upcoming lending protocol, when it's live. 7. Build AI apps on @flop_labs during testnet. -> Vibe code your own apps. Use free testnet inference to get your first users and position yourself for the FLOP agent airdrop. The goal: stay consistent and optimistic while positioning myself for potentially huge airdrops. One year. Locked in.

Last reminder. The Flop Network launch event in Singapore is at 9:00AM tomorrow and you will need to RSVP to enter. @CryptoHayes is taking questions live and food and drinks are on us. Tue Oct 6, 9am to noon SGT.

MiMo-V3 is getting a new architecture. The core of it, HySparse2, is out today. Less prefill, a smaller KV cache, better long-context retrieval—and we got all three at once. Compared with MiMo-V2.6's Hybrid SWA architecture: • 5.02× lower prefill FLOPs at 1M tokens • 4.5× smaller KV cache at 1M tokens • Better MRCRv2 and RULER-v2 scores, plus lower AgentPPL and LongPPL Why build a new architecture? Agentic inference is a very different workload. Each round, a short action can return a long observation that needs to be prefilled, while the context keeps growing. That puts prefill cost, KV-cache size, and retrieval accuracy on the critical path at the same time. HySparse2 tackles all three with two levels of KV sharing: • KV Bridging: Following YOCO, full-attention layers in the cross-decoder build their K/V from self-decoder hidden states. • KV Reuse: Within each hybrid block, sparse layers reuse the preceding full-attention layer's KV cache and selection indices. Two more changes: token-level selection replaces block-level selection, and a forced window of recent tokens replaces the separate SWA branch, so local and global tokens share one KV cache. Since all cross-decoder KV caches now come from the self-decoder, prefill can stop once the self-decoder finishes. Paper: