OpenAI reveals how much its researchers are spending on AI coding
THE SO WHAT
Top OpenAI researchers burning $7,000 per day on coding agents shows how quickly token spend can dwarf traditional dev-tool budgets when teams lean in. If you’re rolling out AI coding internally, treat usage caps, cost dashboards, and ROI tracking as first-class product requirements, not afterthoughts.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIByteDance is preparing a real-time spatial video model under Zhang Yiming, Bloomberg reports
A ByteDance world model for real-time spatial video, personally overseen by Zhang Yiming, is a bet that 3D generative understanding will be a core primitive for both content and robotics. If you’re building in AR, telepresence, or autonomy, assume Chinese consumer-scale data will soon be training competitive spatial models and plan your data and partnership strategy accordingly.
Applied AIImport AI 472: DeepMind's cheating math agents; populist AI policies; and Forethought theorizes a nightwatchman
The emerging theme across cheating agents, populist AI policy, and "nightwatchman" ideas is that governance is shifting from model weights to usage patterns and institutional controls. Leaders should be logging not just what models can do, but where incentives might push them—and users—to exploit loopholes.
Applied AIByteDance Founder Joins AI Elite in Race to Perfect World Models
ByteDance entering real-time spatial world models alongside Meta and Alphabet means this is now a three-way race with consumer-scale data on all sides. If your roadmap depends on proprietary 3D understanding, assume the baseline capability will commoditize and focus on proprietary data, hardware integration, or vertical workflows.
Applied AI‘Other governments will be watching this closely’ — South Korea starts testing free AI for the public
A state-backed, free public AI service in South Korea is a live test of AI as digital public infrastructure rather than a pure commercial product. For operators, this raises the bar on UX and trust—citizens may start benchmarking private assistants against whatever their government provides by default.