The AI 2040 plan calls for pausing frontier training runs via inference-only verification on clusters >10k H100e, mandatory compute inventories, physical chip counts, and air-gapped R&D fac…

The AI 2040 plan advocates pausing frontier training runs via compute governance — requiring inference-only verification for >10k H100 clusters, mandatory compute inventories, physical chip…

The AI 2040 proposal calls for pausing frontier training runs, requiring inference-only verification for data centers above 10k H100 equivalents, mandatory compute inventories, foreign insp…

The AI 2040 proposal calls for pausing frontier training runs via inference-only verification at data centers with >10k H100 equivalents, mandatory compute inventories, foreign chip inspect…

The AI 2040 plan calls for pausing frontier training runs via inference-only verification at data centers with >10k H100 equivalents, mandatory compute inventories, physical chip counts, an…

The AI 2040 proposal explicitly permits continued inference on current models (Astra, Fable, Grok) and capabilities research, but pauses new frontier training runs via compute caps and veri…

Concrete slowdown framework allows current inference to continue while gating next-gen training behind government-verified compute inventories, chip tracking, and bandwidth-capped (1 Mbps)…

The AI 2040 proposal advocates pausing frontier training runs via inference-only verification at data centers >10k H100e, mandatory compute inventories, physical chip counts, and air-gapped…

The AI 2040 framework requires frontier model weights moving from R&D to inference facilities to be stored on physical devices encrypted independently by both US and China, with representat…

A fruit-fly connectome mapped by Google is now simulated in software and connected to a Hugging Face SO-100 arm via OpenAI's Astra/Codex, enabling painting, Doom, and parallel parking — sig…

Hugging Face's SO-100 open-source 3D-printable robot arm (~$200 actuators) combined with frontier models (Astra) enables low-cost physical AI experimentation; YC humanoid under $2K and Nat…

Whole-brain emulation of a fruit fly (via Google connectome) now runs in simulation and drives a $200 Hugging Face SO-100 arm, signaling a 'deep research moment' for robotics where cheap ha…

Cheap open-source robot arms (Hugging Face SO-100) combined with vision-language models like Astra enable rapid iteration on physical tasks — painting, mail sorting, weed pulling — lowering…

Hugging Face's SO-100 arm (~$200) and sub-$2K humanoids from YC companies, combined with models like Astra that can plan and execute physical tasks via Codex, lower the stakes for experimen…

Ty shows that current vision-language models can plan and execute physical painting tasks on a low-cost open-source robot arm, suggesting a near-term 'deep research moment' for robotics whe…

Cheap commodity hardware + frontier vision-language models enable rapid robotics prototyping; the 'deep research moment for robotics' is arriving as agents close the IRL-to-digital loop (ma…

Guy Siri applies A&R intuition — acting fast on 'chorus' moments, ignoring herd noise with 'blinders,' and trusting gut over consensus — to identify generational AI platforms, acknowledging…

Guy Siri and Alex describe venture as identical to A&R: maniacal focus on niche, blinders against herd noise, gut-driven conviction honed by pattern matching across cycles, and willingness…

Early conviction in foundational labs (near $1B deployed) created a high watermark; today's crowded landscape requires more work and pattern recognition to distinguish signal from noise, bu…

Guy Siri translates music talent-spotting — identifying the 'chorus' in a founder's pitch, acting fast before competitors overpay, trusting gut over consensus — to venture, citing early Ope…

Guy Siri translates music talent scouting (signing Alanis Morissette, Muse on first song) to venture: maniacal focus on founder 'chorus', ignoring market noise via 'horse blinders' (David G…

Personal agents (Instinct, Town, Muse, Meta) will compete on network effects where agents can port user contacts across platforms, and monetize via take rates on agentic commerce (e.g., 500…

Alex argues that widespread AI anxiety drives users to try new agent products to avoid obsolescence, giving startups a unique forgiveness window and network effect opportunity against incum…

Alex argues fear of being left behind drives users to try new agent products despite existing daily drivers, creating a unique window for startups like Instinct to challenge incumbents — ne…

Investors see personal agents as a massive new category where fear of obsolescence drives multi-product trial, enabling challenger startups (Instinct, Town) to build network effects via con…

Positron's Atlas chip uses LPDDR5X instead of HBM to avoid CoWoS constraints, achieving competitive inference TCO and interactivity; the team shipped FPGA-based Gen1 in 15 months with <20 p…

Positron AI uses FPGA-based first-gen and custom silicon second-gen (Atlas racks at Oracle) to run frontier models on commodity LPDDR5X memory, bypassing HBM/CoWoS supply constraints; they…

Neoclouds (Lambda, Fluid Stack) provide GPU cloud infrastructure; Positron's Atlas FPGA racks deployed at Oracle Cloud show new silicon vendors leveraging neocloud channels for early revenu…

Google's VO video model required 4 H100s for 10-second clips, completely memory-bandwidth and capacity bound; Positron's architecture uses commodity LPDDR5X to solve the memory bottleneck,…

Video generation models are memory-bound on capacity and bandwidth; solving this with commodity memory architectures (LPDDR5X) enables new silicon entrants to scale without HBM/CoWoS constr…

Mitesh Agarwal observed Google's Veo video model required 4 H100s for a 10-second clip and was 'completely memory bound on bandwidth and capacity,' motivating Positron's commodity-memory ar…

Video and reasoning models are memory-bandwidth and capacity bound (Google VO needs 4 H100s for 10s clip); Positron's architecture uses commodity LPDDR5X to bypass HBM supply constraints an…

Mitesh Agarwal explains that inference is memory-bound on bandwidth and capacity, so Positron uses commodity LPDDR5X instead of HBM to bypass CoWoS packaging constraints, enabling faster sc…

Nvidia's annual silicon refresh (Blackwell, Rubin, etc.) sets a pace that any competitor must match to stay relevant; Positron's 15-month FPGA-to-customer timeline and 2027 ASIC tapeout ref…

Hyperscalers demand proof of fab capacity (hundreds of MW by 2028, GW by 2030) before committing; Positron's $875M raise and VTA partnership aim to satisfy TSMC and customer scale requireme…

Mitesh Agarwal argues that AI chip companies will find market adoption if they can make their silicon run frontier models efficiently, as hyperscalers prioritize total cost of ownership and…

Positron's architecture uses commodity LPDDR5X memory instead of HBM to avoid the CoWoS packaging bottleneck that constrains Nvidia, AMD, and TPU supply; this enables faster scaling but req…

Nvidia's 12-month silicon cadence sets the pace; Gavin Baker argues 1% of the AI chip TAM equals $100B enterprise value (implying $10T market), driving $9B into silicon startups in 12 month…

Frontier labs and hyperscalers require chip vendors to show credible paths to hundreds of megawatts by 2028 and gigawatt-scale by 2032; Microsoft's 10GW/26GW/38GW expansion illustrates the…

Frontier labs and hyperscalers now require silicon vendors to show credible paths to hundreds of megawatts by 2028 and gigawatts by 2032, forcing massive upfront capital ($875M raise) and s…

Frontier labs and hyperscalers now require chip vendors to show credible paths to hundreds of megawatts by 2028 and gigawatts by 2032, forcing startups like Positron to raise large equity r…

Frontier labs and hyperscalers evaluate new chips on two criteria: ability to fabricate at gigawatt-scale (hundreds of megawatts by 2028, gigawatts by 2030) and deployability within existin…