OpenAI ends fine-tuning services as AI elite pursues alternative strategies; fine-tuning dies as mainstream approach while top applications favor open-source models and long-prompts, with agent-based systems breaking new benchmarks in research and mathematics.
OpenAI releases GPT-Realtime-2 with 15.2 percent performance improvement; new features include configurable preambles, parallel tool calls, better error handling, expanded context length to 128K tokens, and improved voice nuance control for developers.
Anthropic secures access to the Colossus I supercomputer through a SpaceX deal worth five billion dollars annually, achieves 8000% annualized ARR growth, and introduces 200+ new AI features, with founder Dario Amodei predicting individuals will build billion-dollar companies in 2026.
Anthropic and OpenAI are establishing specialized services divisions to integrate AI systems tailored to enterprise processes and unlock new revenue streams.
The AI Engineer World’s Fair opens its Wave 2 speaker submission call, expands its program with six new topic tracks, and expects to host the year’s largest technical AI conference in San Francisco.
With Physical AI, model intelligence is not the main bottleneck; rather, it is safe, reliable deployment on hardware under latency, energy, and cost constraints.
