You changed the incubation time. Or the buffer. Or the order of the wash. Your partner is coming in tomorrow and will run the old method unless you tell them — and "I tweaked a few things, I'll explain in person" is how you get two datasets that cannot be compared.
This is not a status chat. It is a confirmation. What changed, why, and the one thing they must agree before the next run.
Write the delta, not the evening
The reader needs the difference between the protocol they last used and the one you want them to use now. Three lines usually cover it: the step, the old value, the current value. "Incubation is 20 minutes at room temperature, not 10 on ice. I changed it after the pellet stayed soft on Tuesday."
Do not send a narrative of every tube. Do not paste the assistant's recap of your troubleshooting session. They were not there, and a cold reader cannot sort your evening into a method.
If you guessed — the concentration came from a bottle with a smeared label, the temperature is what the room felt like — state that assumption. A labelled guess is useful. An inferred number that reads as measured is how the next run fails for a reason nobody can find.
Ask them to confirm one thing
Ask for a decision: run the changed protocol tomorrow, or hold and talk first. Closed list. If you need them to check a calculation, make that the ask and drop the rest.
A sufficient reply is one line. "Confirmed, I'll run 20 minutes." "Hold — I think the kit insert still says 10." You are not asking for thoughts on the science. You are asking whether tomorrow's work uses the same method.
Name the default. "If I do not hear by 9pm, I will assume we hold and I will not start the incubation." Silence is otherwise read as assent, and assent you invented is worse than a delay.
Keep inferred claims labelled
Your assistant will fill gaps. It will write a tidy methods paragraph that includes a volume you never measured. Provenance labels exist so your partner can see which sentences you stand behind and which the model inferred. Read the draft as if you were the one at the bench. Cut anything you would not write on the notebook page.
You approve the send. Draft, then confirm. The package does not go out because the model sounded sure.
If your partner has no account, they get email and a link. If they already use RelayLink, you need an accepted contact. Nicknames stay private unless you share them — write "Priya," not the joke name you typed into your own list.
No attachments. The shared notebook, the kit PDF, the course protocol live where they already live. The briefing is the change and the confirmation.
A block, if they ever use one, is silent. Do not treat a missing reply as a mystery when you already named a default.
When the delta is short enough to confirm from a phone, connect your assistant and send the change before they thaw the samples.