The AI news that matters,
in two minutes a day.

Every morning we pick the AI stories people are actually talking about and explain each one in a short, sourced briefing. No hype, no jargon walls.

Oct 16, 2026

1 min read

Meta released Muse Spark Open, an 816B-parameter open-weight multimodal MoE model that runs locally on two M5 Ultra Mac Studios via Apple MLX.

Open models

1 min read

StepFun released open weights for Step 5 Preview, a 600B-parameter sparse MoE model with 27B active parameters and a 1M-token context window.

Open modelsAgents

1 min read

NVIDIA launched Nemotron 3.5 Super VL, a 120B open multimodal MoE, plus Bolt 74B-A9B, which fits in 24GB for local agents on RTX and DGX Spark.

Open modelsResearch

1 min read

Microsoft AI finished pre-training MAI-Base-2, a 2.8T-parameter MoE trained on 16,000 NVIDIA GB200 GPUs in 38 days. Post-training is underway.

ComputeIndustry

1 min read

Datadog's Scout 1.0 post-trains GLM-5.3 Flash with reinforcement learning for incident investigation, matching frontier models at 3-8x lower cost.

ResearchIndustry