"Frontier": A Word That's Traveling Too Fast
This month, I heard half a dozen organizations introduce themselves as "frontier". None of them trains a model. That deserves a closer look.
Read article →EPYCREST / BLOG
AI, in practice.
Analysis, field experience and methods for building software with artificial intelligence.
The Open Bar Is Closing: Thinking in Tokens, Not Just Hours
The era when you could throw any model at any task is over. The new skill to develop: choose the right model, give it the right context, and learn to reason in tokens as much as in hours.
Read article →
That's So February 2026
In a world where AI redefines our practices every week, what was innovative two months ago is already outdated. Our approaches, tools, and instincts must evolve at the same pace — and the people best equipped to help us stay on course aren't always who you'd expect.
Read article →Local or Cloud AI? The Real Math Nobody's Doing
Your data leaves your infrastructure with every AI request. Your cloud bill climbs without warning. What if the answer was already in your server room? The complete guide to choosing between local inference and cloud APIs in 2026.
Read article →
AI and White-Collar Work: The Trust Problem Nobody Names
AI can already perform many white-collar tasks. But organizations skipping the governance and trust-building step are taking a structural risk that is still widely underestimated.
Read article →