Why this course is a snapshot, not a permanent map
Everything covered in this course — which labs are making which bets, what agents can reliably do, how the capital and bubble debate is shaped, where open-weight sits relative to closed — is an honest description of the landscape as of 2026, and every part of it is subject to change, some of it quickly. This isn't a disclaimer to skim past; it's the single most important thing to internalize about following AI as a field. A new model release can shift the open-weight/closed capability gap in a matter of days. A single high-profile agentic failure or success can swing the 'agents actually work' narrative in either direction. Funding announcements, restructurings, and new lab strategies happen on a timescale that would make any 'current as of' snapshot embarrassing within a year if treated as permanent truth.
The right response to this isn't anxiety about needing to constantly relearn everything — it's recognizing that the durable parts of what you've learned (the different strategic bets labs make, the bounded-scope-plus-fast-feedback heuristic for real AI value, the open-vs-closed tradeoff structure, the fact that serious people genuinely disagree about the bubble question) are frameworks, not facts, and frameworks age much more slowly than facts do. A framework like 'ask what a lab is betting on, not who's currently ahead' stays useful even as the specific answer to 'who's ahead' changes monthly.