What does it mean that the minimum viable behavior change?

What is the smallest change you could make to test whether this approach works.

Example: You want to become a person who writes daily. The ambitious version is a two-hour morning writing session with a thousand-word minimum before you check email. The MVBC is opening a blank document at your desk after your first cup of coffee and writing one sentence about whatever is on your mind. One sentence. Not a paragraph, not a page, not a "real" writing session. One sentence. On day one you write a sentence about a project frustration and stop. On day two you write a sentence and then, without planning to, write three more. On day four you notice you are writing for twelve minutes without having decided to write for twelve minutes. The sentence was not the writing practice. It was the behavioral kernel — the smallest unit that still produced a testable signal. The signal it produced was clear: you are willing to open the document, you do have something to say, and the friction is in starting, not in continuing. That signal tells you the full practice is viable. A two-hour commitment on day one would have told you nothing except whether you could white-knuckle through a single morning.

Try this: Choose a behavior change you have been considering, one you have either not started or have started and abandoned. Now strip it down to its behavioral kernel by asking three questions. First, what is the core action — the single irreducible physical or cognitive behavior at the heart of this change? Second, what is the smallest version of that core action you could perform that would still produce observable data about whether the full version is worth pursuing? Third, what specific signal would you look for to distinguish "this has potential" from "this is not working"? Write your MVBC in this format: "For five days, I will [one specific micro-behavior] in [one specific context], and I will observe [one specific signal]." Run the MVBC for five days. On day six, review: did the signal appear? Was the behavior sustainable at this scale? What surprised you? Based on those answers, decide: scale up by one increment, modify and retest, or abandon and try a different behavioral kernel.