Qwen3.8-Flash-Next

• <p><strong><a href="https://qwen.ai/blog?id=qwen3.8-flash-next">Qwen3.8-Flash-Next</a></strong></p> Another open weights model from Qwen. This one is "a multimodal MoE model that also serves as an early preview of the architecture used in Qwen4".</p> <p>It's pretty big: 125B tokens, but only 6B acti

OpenAI ships its own inference chip while Anthropic and Perplexity race to own the 'memory + local execution' layer of AI agents.

• OpenAI unveiled Jalapeño, a custom inference chip co-built with Broadcom in 9 months (partly AI-designed), targeting ~50% lower cost per response than Nvidia chips — infrastructure-only, not for rent/sale.
• Anthropic made unified memory default across Claude chat and Cowork agents on Free/Pro/Max plans — context persists across sessions, with manual edit/delete controls and opt-in for sensitive topics.
• Perplexity launched Portable Computer, a fully local agent (orchestrator + subagent + tools) running on NVIDIA DGX Spark hardware — zero token cost for local tasks, opt-in cloud escalation, available now for Pro/Max on Linux.
• Prime Intellect open-sourced an agent harness that boosted ARC-AGI-3 scores from 30% to 95.5% by enabling context reuse across runs.
• OpenAI also launched a $100/seat ChatGPT Business Premium tier and a $35K hackathon for WebMCP, its open standard for making websites agent-ready.

Apple's new desktop computers are designed specifically for local AI development

Quantization-Aware Healing: a compressed, 4-bit model that outperforms its full-precision original

Google announces Gemini 3.5 Transcribe for AI-powered speech-to-text

Apple, OpenAI, and Nvidia/Perplexity are all racing to make local AI inference cheaper and faster, signaling a real shift away from cloud-only agent workflows.

• Apple launched M6 Mac mini ($899) and M5 Ultra Mac Studio ($5,499), claiming up to 4x and 4.3x faster AI performance with Neural Accelerators built into every GPU core, explicitly marketed for local AI workloads and even Mac Studio clustering for inference.
• OpenAI's first custom inference chip, Jalapeno (built with Broadcom), beat Nvidia Blackwell by 1.5-4x on performance-per-watt in independent SemiAnalysis benchmarks, developed in under two years partly by using OpenAI's own models to speed up chip design.
• Perplexity launched 'Portable Computer,' a local AI agent running on Nvidia's DGX Spark hardware using open models (PPLX 27B based on Qwen), aimed at cutting token costs, boosting speed, and keeping sensitive data private.
• Mac mini shortages earlier in 2026 were driven largely by AI enthusiasts running agents like OpenClaw as always-on local machines, confirming real demand for local inference hardware.
• MCP protocol received its biggest spec update since June 2025 (July 2026 revision): sessions removed, DCR deprecated in favor of CIMD, and six new authorization SEPs — relevant for anyone running MCP servers for agent tooling.
• OpenAI plans to ramp Jalapeno volume into 2027, targeting latency-sensitive products like a faster Codex mode rather than replacing Nvidia chips broadly.

IBM's new Granite 4.2 models ride the wave of interest in local LLMs

Anthropic polishes Claude's UX infrastructure (renderer + enterprise auth) while Pi's disk-spill trick slashes agent context costs by up to 88%.

• Anthropic rebuilt Claude's streaming renderer to only update actively-changing text instead of re-rendering full replies, cutting worst-case freezes 4.5x and holding steady 120fps on MacBooks.
• Anthropic shipped enterprise-managed auth for MCP connectors, removing manual OAuth setup for organizations connecting Claude to internal tools.
• Pi's coding agent uses a 'prune + spill' memory pattern — full tool output is saved to disk while context holds only a file path — cutting context usage 26-35% and processing costs up to 88% across 19 real sessions.
• An open-source Claude-based tool (built by a laid-off geophysicist) uses a drafter-reviewer agent pattern to auto-write tailored CVs and cover letters; landed 20 interviews from 69 applications.
• VoroTracing renders 3D scenes at 623 FPS, beating Gaussian Splatting by 2.8x, signaling a real-time rendering stack upgrade.
• Thinking Machines launched $50,000 in grants for open-weight AI safety research.

OpenAI's price war and Thomson Reuters' proprietary LLM show enterprises are getting disciplined about what AI intelligence is actually worth.

• OpenAI cut GPT-5.6 Sol prices over 20% (now $4/$20 per million input/output tokens), undercutting Anthropic's Fable 5 ($10/$50) and even Opus 5 ($5/$25).
• Ramp payments data shows enterprises are skipping top-tier models: Fable 5 is only 6% of Anthropic token spend, while cheaper Opus 5 has overtaken it in corporate usage.
• Thomson Reuters launched 'Thomson,' a proprietary LLM built on an open-source base plus decades of Westlaw/Reuters/Practical Law data, for $40M—far below frontier training costs—now live in CoCounsel Legal.
• Indeed's new AI exposure index ranks San Jose (59), Seattle (57), Washington DC (54), San Francisco (53), and Austin (52) as most AI-exposed metros, driven by software/knowledge work; trade and hands-on jobs remain resilient.
• Nvidia is putting $6B into a US open-source AI ecosystem and reportedly negotiating a Perplexity investment at a $30B valuation.

Higgsfield Global Film Festival: How to Enter, Rules, and the $1,000,000 Prize Pool

• Higgsfield Global Film Festival is now open for submissions. It's a $1,000,000 short film contest made entirely on Higgsfield's platform. Any genre, any style, minimum 3 minutes, solo or in a team of up to four. This guide covers who can enter, how films get made, and how prizes and rights work.

Granite 4.2 LLMs: How They're Built

AI agents meant to replace Meta workers made “large-scale, disruptive actions”

Quoting Paul Dix

• <blockquote cite="https://pauldix.com/the-end-of-programming"><p>The fact that AI wrote 1M LOC and then refined it over the course of the next couple of months to produce a reliable piece of software that is currently running on millions of developer machines is absolutely mind blowing. And you can