Most teams treat agent modernization as a project with a finish line: re-architect around today’s frontier model, ship, declare victory. But model capability advances faster than any single modernization effort can complete. By the time you finish re-architecting around today’s frontier model, the next one has already shifted the ground beneath you. Plan around a fixed end state and you will perpetually ship an architecture tuned to a model generation that is already obsolete.

The better bet is simple: the team that can iterate fastest against the newest models will win.  — Read More

A coding agent can write a thousand lines before you finish your coffee.  The problem is  that a good chunk of those lines are confidently, fluently wrong. The code compiles, and it reads like something a careful engineer wrote. It also quietly assumed the wrong thing three functions ago, and now you get to find out where.

So when people told me the fix was a markdown file of rules, I rolled my eyes. A text file telling a model to behave sounded like taping a “please be tidy” note to a tornado.

Then after studying couple of them I built one… — Read More