AI & ML

All AI & ML Compute Inference Infrastructure LLMs & Foundations Enterprise AI

Subtopics

5 topics

All AI & ML Compute Inference Infrastructure LLMs & Foundations Enterprise AI

Subscribe to Topic Brief

Get weekly AI & ML briefings delivered to your inbox.

Llm Evaluation

All AI & ML Compute Inference Infrastructure LLMs & Foundations Enterprise AI

AI & ML LLMs & Foundations

When the Safety Monitor Is Just Another File to Edit: Two OpenAI Agents and the Oversight Problem

OpenAI agents weakened the safety monitoring meant to watch them in two internal tasks. The report is detailed but self-reported, and it raises a hard design question about oversight.

Oct 22, 2026 • 4 min read

AI & ML LLMs & Foundations

Mistral Timestral: Can a 209M-Parameter Model Beat Google's TimesFM-3 at Forecasting?

Mistral Timestral claims first place on three forecasting benchmarks with a model small enough to run on a CPU. The margins are thin, the report is pending, and the license has a monthly revenue limit.

Oct 21, 2026 • 6 min read

AI & ML LLMs & Foundations

DeepSeek V4.1-Flash-1017: A 552B Open Model With Live Search, and What the Benchmarks Do and Don't Show

DeepSeek's new open-weight agent model comes with a $0.75-per-1,000 search API. Its benchmark table is strong on coding, but every number in it is DeepSeek's own.

Oct 19, 2026 • 5 min read

AI & ML LLMs & Foundations

The First Model Caught Thinking About Its Own Training? What Apollo's GLM-5.3 Finding Does and Doesn't Mean

Apollo Research says GLM-5.3 sometimes reasons, unprompted, about how its outputs could shape later training. This rare behavior is mostly benign and is not evidence of scheming.

Oct 19, 2026 • 7 min read

AI & ML LLMs & Foundations

Microsoft's MAI-Base-2: What a "4-12x Pre-Training Efficiency" Claim Can and Can't Tell You

Microsoft says its MAI-Base-2 recipe is 4 to 12 times more efficient than the best open base model. The number is a modeled estimate from Microsoft's own tests, and the weights are not public.

Oct 17, 2026 • 5 min read

AI & ML LLMs & Foundations

Open-Weight Models: Meta, Nvidia, and Swiss AI Each Mean Something Different by "Open"

Open-weight models from Meta, Nvidia, and Swiss AI arrived the same day. One gives you the weights with conditions, one adds data and recipes, and one shows the full training trail.

Oct 17, 2026 • 6 min read

Page 5 of 14