Hello friends — in this livestream I’m breaking down what *actually* changed from 2025 into early 2026 that makes AI agents way more powerful. Yes, the flagship models got better (more context, better coding, better tool use, more autonomy), but the singular biggest shift is the open source explosion. In the past few months, open source models went from “nice” to legitimately competitive with closed source, and that changes how you should think about building agentic systems if you care about cost, privacy, control, and not getting locked into one vendor. I walk through why open source matters in practice: full control over your data (especially for regulated industries), customizability, auditability, and huge cost reductions when you’re running high-volume workloads. I also keep it real about the tradeoffs: setup is a bit more involved, hardware matters (my MacBook Air was suffering when I tried to run too many models while livestreaming), and you’re responsible for managing your own infra. Then I zoom out to the fundamentals: agentic systems still come down to models + tools + knowledge/memory + audio/speech + guardrails + orchestration. The hype changes, but those components don’t — and if you internalize that, you’ll stop freaking out every time a new framework drops.