The honeymoon phase of "just deploy an AI agent" is officially over. Today's reality check: agents need babysitters, vendors are becoming service providers, and the infrastructure around AI is getting more complex, not simpler.
01
Box CEO Aaron Levie: AI agents aren't plug-and-play anymore
Levie points out that selling AI agents to enterprises has fundamentally changed the vendor relationship. You're no longer just selling software that gets implemented once. You're selling an ongoing workflow that requires constant management and iteration. The "forward deployed engineer" role isn't disappearing — it's becoming more critical as companies realize agents need dedicated oversight.
Why it matters: That AI vendor promising to "automate everything with one integration" is about to become your new consulting partner. Budget accordingly.
Peter Yang shares the "AI is clay, not a factory" philosophy
Yang highlighted advice from an unnamed expert: treating AI output as a final result is a mistake. Instead, you should think of AI as giving you raw material to shape. The real skill is knowing what's worth exploring from AI's "almost infinite" possibilities, not just accepting whatever the prompt returns.
Why it matters: Companies hiring "prompt engineers" are solving the wrong problem. The valuable people are the ones who can judge and refine AI output, not just generate it.
Y Combinator's Garry Tan details his personal AI knowledge base
Tan revealed that his personal GBrain system now contains over 17,000 pages and was built in 12 days using his open-source GStack toolkit. The system is free and MIT licensed. His emphasis: "It knows me" — suggesting the value is in personalization, not just document storage.
Why it matters: Personal AI assistants that actually know your work patterns and preferences are moving from concept to reality. Expect more tools focused on learning your specific context, not just answering generic questions.
Josh Woodward from Google announced NEET practice tests in Gemini, targeting India's largest medical entrance exam. The company is asking users what subjects and countries to add next.
Why it matters: Google is moving beyond general chat into specific, high-stakes use cases where accuracy matters more than creativity.