We announced the Echo OS pilot with an essay about why it is shaped the way it is. This is the other half: what to do with the grant once it lands.

It is written for someone who has applied, or is about to, and wants to know what a good ninety days looks like.

What you actually have

$250 of your twin’s work, for ninety days, spent before anything else you have paid for. The pilot page has the terms in full.

The useful way to hold it is this: it is not a discount on a bill. It is a budget for work.

A discount makes something you were already buying cheaper. A budget asks a different question — what should this be spent on? That question has a right answer and several wrong ones, which is the reason this essay exists.

Spend it on work that already exists

The instinct with a new tool is to invent a project for it. Resist that one. A project invented for the grant tells you how the twin handles a thing you made up.

Point it instead at work that is already happening and already imperfect:

  • The question customers ask most often, that you answer identically every time.
  • The calls that arrive when there is nobody free to take them.
  • The follow-up that everyone agrees matters and nobody gets to by Friday.

These are unglamorous, which is the point. They are also the only work you already have a feel for, so you will know what you are looking at without needing a dashboard to tell you.

$250 is deliberately not a lot. It is enough to find out whether the twin can hold a real piece of your operation, and not enough to disappear into an experiment. Treat the size as a hint about scope.

What it will not pay for, and what that means in practice

The grant pays for the work your twin does. It does not pay for a subscription — plans bill separately, to your card.

In practice that means the two never touch. A paid plan keeps billing to your card right through the pilot, and the grant sits underneath it, covering what the twin actually does. The invoice is one thing and the grant is another, spent where the work happens rather than where the billing does.

There is one more thing worth knowing before day one, and we would rather you read it here than discover it. Pilot credit does not yet appear as its own line in the app. Your balance still reads zero while the credit is working, and the app may still prompt you to add a card. Nothing has gone wrong when that happens. Until that screen exists, we confirm your balance and your expiry date in writing, and that written confirmation is the record.

Ninety days, and what to do with them

Credit expires ninety days after we apply it. Anything still sitting there on day ninety-one is written off: no refund, no rollover, no grace period to ask about.

Ninety days sounds generous and goes quickly. If it helps, a shape that tends to make sense:

The first stretch is for getting it wrong. Set the twin up, watch what it says, correct it. This part costs the least and is worth the most, because everything after it inherits what you fixed here.

The middle stretch is the actual test. Hand over the recurring thing — the one you picked above — and leave it there long enough to see a normal week, including a bad one.

The last stretch is for reading what happened. Not the total spent, but where it went: what the twin was asked for most, what it handled, and what it handed back to you.

Day ninety-one

The unused remainder goes. That is the part people brace for, so it is worth being clear about what does not go.

Your twin stays. What it learned about your business stays. The work it did stays done. What changes is how the next piece of work is paid for: after the grant, it is billed the ordinary way, like any other work you ask for.

An expiring grant is not a trapdoor. It is a boundary around an experiment, drawn so the experiment has to end and be judged.

The next grant

A business may apply up to four times, three months apart. Across the whole pilot that tops out at $1,000.

Two honest notes about that. First, nothing in the app counts this for you — there is no meter, no tracker, no eligibility screen. It is a policy we keep by hand, the same way we grant the credit by hand. Second, every application is reviewed on its own, including the second, third and fourth. Approval is never automatic, and having had one grant is not a claim on the next.

What makes a later application easy to say yes to is not enthusiasm. It is specifics: what you put on the twin, what happened, and what you would hand it next. That is the report we are hoping for, and it is the same report you would want for yourself.

What we get out of it

We should say this plainly, since we are asking you to spend our money thoughtfully.

We learn what businesses genuinely hand to a twin when the cost of trying is not the obstacle. We have opinions about that, formed from building the thing, and opinions formed that way are exactly the ones worth checking against someone else’s Tuesday. A pilot where everyone politely tries the features we expected would teach us nothing.

So spend it on the thing you actually need done, not on a tour of what we built. If it turns out the twin is wrong for that thing, we would rather learn it in ninety days than in a year.

Terms and the way to apply are on the pilot page. If you want the argument behind the terms rather than the practice, that is the announcement.