NESTFRONTIER / ARCHIVE
Technical analysis, model releases, research, and deployment notes.
Microsoft is launching Project Perception, a multi-model AI security tool that uses Anthropic, OpenAI, and Microsoft models to take on Mythos 5 in the gap created by June's export controls.
Four DeepMind researchers left in a week. Gemini 3.5 Pro missed its coding targets. $425 billion erased from Alphabet. The AI talent war just got expensive.
LM Studio launched Bionic, a desktop agent app for open models that handles coding, documents, and voice — all running locally or with Zero Data Retention cloud routing.
Moonshot AI dropped Kimi K3, a 2.8-trillion-parameter open model that matches Claude and GPT on key agentic benchmarks. Open weights arrive July 27.
A security researcher caught Grok Build shipping full Git repositories to xAI servers at 27,800 times the data a coding task required. Then xAI open-sourced it.
Thinking Machines Lab released Inkling: a 975B open-weight MoE model with Apache 2.0 license, controllable thinking effort, and native multimodality across text, images, and audio.
A zero day in Cursor runs arbitrary code from any repository you open. Mindgard reported it seven months ago. Nothing was fixed until they went public.
PrismML's Bonsai 27B compresses a full 27B-class reasoning model to 3.9 GB, running on an iPhone 17 Pro at 11 tok/s. The tradeoffs are real, but so is the trajectory.
A 744B-parameter open-weight model from Beijing lands within a point of Claude Opus 4.8 on agentic benchmarks at one-fifth the cost. The margin mirage is over.
Uber blew its entire 2026 AI budget in four months. Microsoft is canceling Claude Code licenses. The enterprise AI coding cost crisis has arrived.