An example record

One result, and everything still attached to it.

This is what the thing we keep going on about actually looks like. Not a screenshot of an interface — the record itself, and what sits underneath each part of it. The study is invented; the structure is not.

This record
An example, made up for this page
The structure
Real — it is what woodle holds
The interface
Not shown. It is being built

Start at the sentence a paper would carry.

a result, with everything still attached · example

Aggregate formation fell 41% at 10 µM, across three plates run together.

Who ran it
Your student, three years ago
When
14 March, morning
On what
The spectrophotometer in room 214
Signed off by
You, on 2 April
The file
plate-read-14-03.csv · 41 KB
Written by
The instrument, unedited since
Checked
Same checksum as the day it landed

Three steps, and the rule down the left grows as you go. That is the whole idea, drawn small: a conclusion reaching back to the thing it rests on.

What each layer is actually for

Most notebooks keep the first row well. The question that arrives years later is about the third.

The claim
A sentence somebody was willing to put their name to. It is the only part that ends up in a paper, and on its own it is unfalsifiable — which is why the rest exists.
The run
Who, when, on which instrument, under which protocol version. This is the row that is usually in somebody's memory, and memory is what leaves when they do.
The file
What the instrument wrote, byte for byte, where it already lives. Not a copy pasted into a notebook — the original, still in your own storage.
The signature
A person deciding this is ready to leave the lab. The assistant can draft the sentence; it cannot do this, on any plan. Where the assistant stops →

Now the part that matters: it was wrong.

Eleven months later somebody notices the blank correction was applied to the wrong column. In most systems this is where the record quietly stops being trustworthy, because the fix overwrites the thing being fixed.

the same result, after a correction · example

Aggregate formation fell 34% at 10 µM, across three plates run together.

Corrected by
Your postdoc, 9 February
Why
Blank subtraction applied to the wrong column
The earlier figure
41% — still here, not deleted
The file underneath
Unchanged. It was never the thing that was wrong
Corrections add, they never overwrite. Both numbers are in the record, in order, with the reason between them — so the history is a sequence rather than a claim about one.

This is the part people underestimate until it happens to them. A record that can be silently edited cannot demonstrate it wasn't — and the problem is not that anybody in your lab would falsify anything. It is that when the question comes, a clean number with no history behind it is indistinguishable from a number somebody tidied up.

Three years on, the question arrives

A reviewer asks which plates, which run, which file. Here is the whole exchange.

01

Open the claim

Search the sentence, or the figure it sits under. You do not need to remember the study name, the date, or who ran it.

02

Follow it down

Run, then file. Both attached, both with their history. The correction is visible rather than hidden, which is the answer to a different question they were about to ask.

03

Send what they asked for

The original instrument file, with the run it belongs to. Not a screenshot, not a retyped table — the thing itself.

What this replaces

Two weeks of somebody going through drives, a folder named by date, and an email to a person who left in 2024. The guide for when that has already happened →

That is the shape. The rest is your lab's.

If it is the shape you have been trying to keep by hand, that is the conversation worth having.

for verified researchersFour questions about your lab. We are setting up founding labs in small waves, by hand.
where we arePaid plans aren’t open yet — the gateway isn’t onboarded, and we’d rather say so than take a card we can’t charge. What else we won’t claim →