When NOT to start an arm

A parallel research arm is not always available to you. The previous lessons describe what an arm looks like when conditions are right. This one is the gating function: three conditions that disqualify the move, and a calendar moment from my own portfolio where I would have failed all three.

If you can't answer past these three, don't start. Park the idea on paper and revisit it in a quarter.

Three disqualifying conditions

A second project is a distraction, not an arm, if any of the following are true on the day you'd start.

1. Your main work is in a launch crunch or burn-the-runway moment. An arm pulls hours from evenings and weekends. In a launch crunch, those hours are not surplus. They are the buffer the main work survives on. Starting an arm in that window steals from the wrong column on the balance sheet. The cost shows up two weeks later as a missed deadline you tell yourself was unrelated.

2. You don't have an honest non-commercial intent. If, when you describe the project to a friend, you find yourself reaching for words like "and if it goes well…" or "the eventual play is…", the project is not an arm. It is a stealth startup. Stealth startups are fine. They are also a different decision with different rails, different time costs, and different consequences for the main work. Mislabeling one as the other is how the arm eats the company that birthed it.

3. You can't name an adjacent topic that compounds. The whole reason an arm is sustainable next to a startup is that the skills compound back into the main work. Sign-language CV shares pose, landmark, and temporal models with cornhole CV; the skill compound is real and named. If you can't articulate a specific, concrete adjacency between the arm and the main work, you don't have an arm. You have a hobby. Hobbies are also fine. They are not what this playbook is about.

The three conditions are not weighted. Any single one is disqualifying. You cannot ace two and squeak through on the third.

My own calendar, told honestly

I started Parley in late April 2026. If I had tried to start it three months earlier, in Q1 2026, every one of the three conditions would have disqualified it.

Q1 2026 was the Mile High Golf launch crunch. Site search was active. The SBA 7(a) loan was mid-process. Grant applications were stacked. Curtis and I were in Hickory most weekends walking sites. There were no evenings or weekends to spend on a Kaggle notebook. The buffer was already spoken for. Condition one would have failed.

I also hadn't yet done the work to write down what Parley wasn't. The early framing of the project, before the "NOT in scope" block went into 12-Parley/project-brief.md, drifted between "side research" and "AR-glasses product on Everysight in 2027." Drift is the symptom of unwritten intent. Condition two would have failed.

The third condition would have passed; sign language and cornhole have shared a CV stack since 2023. But two out of three is enough. The arm would have died inside a month, taken time from MHG, and produced nothing publishable.

Parley started in late April because the calendar opened. The Hickory site got superseded by Denver in April. The SBA package got re-pointed. The first version of the brief got written down with the "NOT in scope" block in it from day one. Three months made the difference between an arm and a distraction.

Three questions to ask before starting

If you are sitting with the urge to start an arm, run the three honest questions. Each has a disqualifying answer.

First: What is the next two-quarter milestone on the main work, and is the path to it boring? Disqualifying answer: anything that includes "we'll know in a few weeks," "we have to get through the next round," or "if X lands." Those describe a window where you cannot afford the diversion.

Second: Can you write down, in one sentence, why this arm will never become a product? Disqualifying answer: any sentence that hedges. "Probably won't," "I don't see it becoming," or "not unless" all fail. The sentence has to be flat. The previous lesson, on the "won't become a product" rule, is the formal version of this check.

Third: Can you name the specific technique or skill that the arm compounds into the main work? Disqualifying answer: a general "I'll learn more about CV" or "it'll sharpen my Python." Those are hobby returns. An arm returns a named technique into a named pipeline. Name both.

If any answer fails, the answer is wait. Write down what would need to change for the answer to flip. Put a date on it.

Demotion path: arms that age out

You start an arm. The conditions were met on day one and then they change. The main work hits a crunch. A new urgent priority lands. A team member leaves and you absorb their work. The honest move is the demotion path.

A demoted arm goes on ice. The repo stays where it is. The decision log gets one entry: "paused on YYYY-MM-DD because of X, resume condition is Y." The repo is not deleted. The work is not erased. It is parked.

Demotion is not failure. It is the discipline that lets you start an arm in the first place. You can begin Parley in April only because you know that if MHG goes red in June, Parley pauses in June without guilt. The kill mechanics, the specific triggers, who pulls the trigger, and how the pause gets communicated to your future self, are the subject of Lesson 8. For now, the principle is enough: the demotion path is what makes the start safe.

Apply this

Pick the arm you've been thinking about starting. Answer the three questions in writing.

If all three pass, you are clear to start. Go to the previous lesson and write the "NOT in scope" block.

If any one fails, write down the specific calendar condition that would need to change for the answer to flip. Put a quarter on it. Park the idea. Revisit on the date you wrote down.

The arm you delay by a quarter is the arm that survives a year. The arm you start under failed conditions is the arm that kills its quarter and produces nothing.