Drew DonaldsonLong context windows limitations, explained: why million-token models still forget, how context rot bites, and when retrieval beats a bigger window.
Drew DonaldsonOpen-weight models for enterprise now route most tokens at a fraction of the cost. When to own your stack, when to rent frontier APIs.
Drew DonaldsonAn anti-hype, anti-doom field report on AI hype vs reality in 2026: where AI quietly earns its keep, where the loud use cases stall, and how to tell them apart.
Drew DonaldsonThe hardest part of agentic AI isn't intelligence, it's integration. Why 26% of agent projects die on legacy plumbing, plus auth, state, and the MCP standard.