Run 125B MoE on a 12GB GPU: Inside Strata’s 45 tok/s Engine
Strata runs 125B Qwen3.8-Flash-Next on 12GB VRAM at 45 tok/s using predictive expert prefetching and io_uring. Here is the PCIe math and setup.
EyesTech: Frontier AI, Silicon & Defence Technology
Independent analysis, benchmarks, and field intelligence on AI models, chips, computing, and defence systems.
SpaceX 800 MHz Deal: The Physics of Satellite Direct-to-Cell
SpaceX acquired 14 MHz of 800 MHz spectrum alongside FCC approval for 15,000 VLEO Gen2 Starlink satellites. Here is the link budget physics for indoor cell.
Microsoft Warns Enterprises Against AI Model Lock-In
At Open Source Summit Europe 2026, Microsoft Director Ryan Waite warned enterprises against proprietary model capture: do not surrender your organizational learning loops to closed APIs. Here is how to architect infrastructure mobility across open-weight models.
M5 Max vs M5 Ultra 512GB: The $10,000 Local LLM Reality
Retailers cut M5 Max Mac Studio prices by $200 as the 512GB M5 Ultra ships at $10,000+. Here is why VRAM residency, not a 26% CPU uplift, dictates the buy.
Drone Strikes Kill Yandex H100 AI Data Centers in Russia
Drone strikes knocked out two Yandex hyperscale AI data centers and 40% of its fleet. Under Western sanctions, ruined H100 clusters cannot be replaced.
Surface Laptop Ultra & RTX Spark: 1 PFLOPS Grace+Blackwell SoC for 120B Local Agents
Microsoft and NVIDIA’s Surface Laptop Ultra pairs an Arm Grace CPU and Blackwell GPU via 300 GB/s NVLink-C2C. Here is what 128GB unified memory and 1 PFLOPS FP4 mean for 120B local agents.
FeSens openTPU: Can AI Agents Design Accelerator Silicon?
FeSens openTPU runs LFM2.5 at 85.8 tok/s on Kintex-7 FPGA silicon via agent-designed RTL, but off-chip DDR bandwidth caps multi-billion models at 1-4 tok/s.
OpenAI’s 10 GW ASIC Alliance: Escaping Nvidia’s 75% Margin
OpenAI and Broadcom booked TSMC 3nm/2nm capacity for a 10 GW inference ASIC deployment by 2029. Here is the CoWoS math, Ethernet fabric, and 60% TCO savings.
Claude Haiku 5.5: Inside the 100K Token Pricing Cliff
Anthropic’s Claude Haiku 5.5 pairs a 1M context with a 5x cost surge at 100k tokens. Here is the worked math on subagent loops, cache reads, and compaction.
Microsoft-Decision-1: 9B Cloud Model in a 35ms Local World
Microsoft-Decision-1 brings 85ms scoring to Azure for $0.042/1M tokens, but WAN transit taxes and 0.4B local models challenge its agent control plane.