Wednesday, February 25, 2026
Top story
Mitchell Hashimoto’s new way of writing code
When the founder of a $15B infrastructure company changes how he codes because of AI agents, it signals a fundamental shift in how technical leaders should be thinking about AI-assisted development workflows.
The briefs
Intercom is shipping foundational improvements to Fin's AI agent architecture, expanding its capability to handle multi-step, complex customer support workflows. This reflects the broader shift from single-turn chat to agentic systems that can reason and execute.
Challenges the prevailing assumption that bigger AI models = better performance. Argues that inference time, not parameter count, is the true bottleneck for model intelligence and learning capacity—reshaping how companies should approach AI scaling.
Mixture of Experts architectures represent a fundamental shift in how models scale compute and parameters. Understanding MoE trade-offs is now table-stakes for any org building or deploying foundation models competitively.
A new product category is emerging around making AI interactions feel more natural and human-like. Companies are raising millions to solve a core UX problem that will determine AI adoption in consumer and enterprise markets.