Microsoft Trained a 2.8 Trillion Parameter Model. The Interesting Part Is How.
A plain-language look at MAI-Base-2: Microsoft's scaling ladder, training efficiency, data provenance, and the model that helped find a pipeline bug.
Read articleWriting
Notes from the work.
Practical AI and cloud engineering, candid developer-relations lessons, and what it takes to build communities that last.
Not Every Decision Needs a Genius. Meet Decision Models.
Meet decision models: fast, low-cost models for routing, classification, and agent control. How Microsoft-Decision-1 and Jev work, and how to try them.
Read article