Discord Goes Ban-Happy, Suspends People for Benign Images
THE SO WHAT
Over-aggressive auto-moderation is the dark side of scaling trust and safety with ML. If your product leans on automated enforcement, you need a fast human appeal path and clear comms — or you’ll burn your power users while trying to protect them.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIOpenAI’s Chief Futurist Is Leaving the Company
Senior safety and strategy departures at major labs are a reminder that the people shaping long-term AI trajectories are mobile. Enterprise buyers should pay more attention to who’s actually steering a lab’s roadmap — and how stable that leadership looks over a 3–5 year horizon.
Applied AITeraWulf CEO Excited About Anthropic Data Center Agreement
A 20-year lease and purpose-built AI campus in Kentucky ties AI infra to long-duration power and land bets. If you’re serious about large-scale training or inference, your real estate and energy strategy now matters as much as your model choice.
Applied AIMemo: SpaceXAI and Cursor plan to launch their first jointly developed AI model as soon as Wednesday, after pushing back the launch to improve model efficiency
SpaceXAI pairing with Cursor to ship a joint model — and delaying for efficiency — underlines how tightly model design is now coupled to downstream dev tooling and infra costs. For builders, the question isn’t just “which model is best” but “which model-tooling combo lets my team ship and run cheapest.”
Applied AIClaude and ChatGPT Are Getting Too Expensive, Even for Microsoft
If Microsoft is routing some Excel and Outlook prompts to its own models on cost grounds, per-token economics are biting even at hyperscale. Enterprises should assume model routing and tiered quality are the new normal — and design workloads to tolerate heterogeneous backends.