AI digest: Agents everywhere, regulation nowhere
NVIDIA drops a compact MoE model, OpenAI consolidates its scattered product line, and Trump's AI framework hands Big Tech exactly what it wanted.
The agent swarm is multiplying. While companies rush to build AI that can actually do things, regulators are stepping back and letting the market sort itself out.
NVIDIA’s Nemotron-Cascade 2 packs intelligence into 3B active parameters
NVIDIA released Nemotron-Cascade 2, a 30B mixture-of-experts model that only activates 3B parameters per request. The model apparently hits “Gold Medal-level performance” while using a fraction of the compute that frontier models need. This feels like the right direction when everyone’s complaining about inference costs.
OpenAI admits it shipped too much, plans desktop superapp
OpenAI is merging ChatGPT, Codex, and Atlas into one desktop application after realising its “ship everything at once” strategy left it exposed. Smart move. Having three separate apps for what should be one workflow never made sense. The real question is whether they can make the unified experience actually better than the sum of its parts.
Trump hands Big Tech the federal preemption it lobbied for
The White House AI framework wants to strip states of their ability to regulate AI, making it a purely federal matter. This is exactly what Big Tech has been pushing for, and it’s probably bad news for anyone who thinks California or New York might want stronger protections. The framework also shifts child safety responsibility to parents rather than platforms.
Pentagon and Anthropic’s messy breakup gets messier
New court filings show the Pentagon told Anthropic they were “nearly aligned” just a week after Trump declared their relationship dead over “unacceptable national security risks”. Anthropic claims the government’s case relies on technical misunderstandings that were never raised during months of negotiations. Sounds like typical government procurement drama, but with higher stakes.