A DeepMind researcher resigned over its AI military deal: 'I couldn't stay at Google in good conscience'
THE SO WHAT
AI talent is starting to vote with their feet on defense work — that turns ethics from an abstract PR topic into a retention and recruiting constraint. If you're touching military, surveillance, or dual-use AI, assume you need a clear internal stance and opt-out paths or you'll bleed your most values-driven researchers.
READ THE SOURCE
MORE FROM THE WIRE
The CEO of a startup helping non-techies code still sees a lot of value in engineering degrees
Low-code that actually works will increase the premium on people who understand systems, not decrease it. If AI can give non-engineers syntax, your differentiator as an engineer becomes architecture, debugging, and owning reliability under load.
Applied AICohere VP says enterprise AI sovereignty requires control of the full agent stack
Enterprise "AI sovereignty" is shifting from model choice to end-to-end agent stack control — orchestration, tools, data, and runtime. If you're serious about sovereignty, you need a reference architecture for what runs where, who owns which layer, and how you can swap components without breaking workflows.
Applied AIThinking Machines Lab Drops Its First Model
A 975B-parameter open model trained for video and audio is another step toward frontier-grade multimodal capabilities outside closed labs. If you're building on open source, start planning for workloads — surveillance, media, robotics — where multimodal understanding, not just text, becomes your core differentiator.
Applied AIThinking Machines amps up its bet against one-size-fits-all AI with its first open model, Inkling
Thinking Machines is using Inkling as a proof point for a "no one-size-fits-all" thesis — big open multimodal backbone plus domain-specific adaptation. For enterprises, the question shifts from "which model" to "what's our customization surface" — data, tools, and constraints layered on top of a shared base.