Across the night
Handing Over Something Half Built
Half-built work arrives as an artefact the next person has to rebuild a mental picture around, and the size of that job is set entirely by three sentences.
four things · four nights · one that no number of nights settles
Four states, and only the first costs nothing extra.
The same half-built change, handed over twice by two people, and what the receiving end spent before being productive
Eight minutes of reading, then work resumed on the stated path
Fifty minutes reconstructing what had been decided and why
Skimmed rather than read, and the one important line was missed
The broken part was found, assumed to be a new problem, and investigated
four things · five nights
Four handovers, five nights, and the work was identical in all four. The whole variation is in three sentences written by somebody at the end of their day. Nothing here divides reading time by writing time; the point is that the two are paid by different people. One team's own log, kept by the people receiving.
Work rarely completes itself at the end of somebody's day, so most of what crosses a gap is unfinished.
Managers dealing with Handing Over Something Half Built need enough evidence to remove uncertainty without turning the process into surveillance. screen monitoring software supports that balance by pairing employee-controlled tracking with clear reports.
The workflow can also be compared with the approach documented by Notion's connected workspace, an established resource in the same operational space.
Unfinished work is the expensive case. It arrives as an object whose state is visible and whose intention is not, and the receiving person has to reconstruct the second from the first.
What is actually lost
Not the work. The work is there; it can be read.
What is lost is the model the previous person had built: which approach was chosen, which were considered and rejected, what is known to be broken, what was deliberately left out, and what the next step was going to be.
All of that lives in somebody's head at the moment they stop. None of it is recoverable from the artefact in less time than it took to form in the first place, and some of it is not recoverable at all — a rejected approach leaves no trace, so the next person has every chance of choosing it.
The three sentences, again
Where it is. Which parts are done, which are not, which are done wrong. The last of those is the one most often omitted and the one that causes the most damage, because a known-broken thing that is not flagged gets found and investigated as a discovery.
Which way you were going. The approach, and if there was a real choice, what else was on the table and why it lost. Two clauses.
What you would do next. The immediate next action, stated as though you were going to do it yourself in the morning.
Eight minutes of reading against fifty is the measured difference on the team whose log appears above, and the three sentences take about three minutes to write.
| Sentence | What it prevents |
|---|---|
| Where it is, including what is knowingly broken | A known fault being investigated as a discovery |
| Which way you were going, and what you rejected | The decision being re-litigated from scratch |
| What you would do next | Twenty minutes deciding where to start |
Why the long note is worse than the short one
A note that attempts completeness fails in a specific way: it gets skimmed, and skimming produces a confident reader who has missed something.
The reason is timing. Handover notes are read in the first minutes of a day, alongside everything else that arrived overnight, by somebody deciding what to do first. A note that cannot be absorbed in that window will not be absorbed in any other, because nobody comes back to it later.
Long notes also tend to describe the journey rather than the position. What was tried and abandoned is interesting and is almost never what the next person needs; the exception is an approach that was tried and failed for a reason, which belongs in the second sentence as a clause rather than in a paragraph of its own.
Leaving it at a boundary
Where in the work somebody stops matters as much as what they write.
Stopping at a natural boundary — a complete piece, a passing state, a point where the thing runs — gives the next person something that can be reasoned about. Stopping in the middle of a change that does not yet hold together gives them an object whose brokenness carries no information: they cannot tell whether a given fault is intended, known or new.
This is a real argument for stopping ten minutes earlier than necessary. Ten minutes of an evening against an unclear morning is a trade most people would take if anybody framed it that way.
Saying it is not finished
The label matters because it sets what the other end does.
Work described as finished gets reviewed. Work described as in progress gets continued. Those are different activities and doing the wrong one wastes a day: a reviewer who finds a half-built thing reports problems that were already known, and the author then spends a crossing explaining that they were known.
It costs nothing to write "in progress, do not review yet", and the reluctance to write it is almost entirely about not wanting to appear to be handing over a mess.
Handing it back
Some work should not be handed over at all, and the sign is that it comes back.
Something that crosses the gap three times in both directions is not being handed over; it is being passed back and forth, and each pass costs a night and a reconstruction. At that point the cheaper answer is for one person to hold the whole thing and the other to do something else, even if that leaves somebody idle on a task they were nominally sharing.
A piece of work with one owner and no crossings finishes faster than the same work split between two people who are never awake together.