… For some models, dropping prior thinking on turns without tool calls is an established behavior and part of the model’s fine-tuning DeepSeek-R1 is the clearest example . But that same behavior is wrong for interleaved agentic turns, where the prior reasoning explains the tool sequence. …
… This looks to be the common theme across tasks from a look at the company’s data, with agentic coding seeing an increase in power because of that better reasoning. …
… Those are then evaluated relative to each other in a “tournament,” the results of which are evaluated by a Reflection agent. An Evolution agent can then make improvements to any surviving ideas, which can be sent back through the process. …
… Having a bunch of AI agents write a bunch of software seems to be a real market for these tools. …
… Having an AI software stack means being able to deploy AI agents into that data to better organize it — with fewer concerns about leakage than there would be with general-purpose third-party LLMs. “Training AI models on public data is the largest, fastest-growing business in history,” Ellison said … …
… "OpenAI unveils AgentKit that lets developers drag and drop to build AI agents" . …