Is this not just journalling?

Journalling is unstructured by design and is not meant to be compared. This is fixed-format specifically so that week five can be read against week one.

What if I had a bad week with nothing shipped?

Then SHIPPED says “nothing”, which is the most useful digest you will get. Two of those in a row is a signal, and the format is built so you cannot hide it from yourself.

Workflow automation

Weekly review digest

One reusable prompt that turns a week of scattered notes into a fixed-format digest, so weeks become comparable instead of blurring together.

Use this once a week, at the same time, with whatever raw material the week left behind.

The pattern

Copy this and replace everything in [SQUARE BRACKETS] with your own specifics.

Prompt template

Here is my raw material from this week: [NOTES, COMMITS, TASKS DONE, MESSAGES, MARKS — whatever exists, unsorted.]

My current goals: [2-3 GOALS, each with how I would know I was making progress.]
Last week's digest, for comparison: [PASTE IT, or write "none".]

Produce the digest in exactly this shape, every week:

1. SHIPPED — what actually got finished. Only finished things.
2. MOVED — what progressed but is not done, with how far.
3. STALLED — what did not move at all this week, and how many weeks it has now been stalled.
4. AGAINST GOALS — for each goal, one line: closer, no change, or further away. Say which evidence you used.
5. PATTERN — one thing this week has in common with last week's digest. If there is no last week, write "no baseline yet".
6. ONE THING — the single highest-leverage thing to do next week, and what it displaces.

Rules: do not congratulate me. Do not pad. If a section is empty, write "nothing" and leave it empty.

Why each part is there

Exactly this shape, every week: The comparability is the entire product. A differently-shaped digest each week is six diaries; the same six sections is a trend you can actually read.
Last week's digest: What upgrades a summary into a review. Week-on-week is where the stall count and the repeating pattern come from.
STALLED with a week count: The most valuable line, and the one no one writes by hand. Three weeks stalled is a decision you need to make, not a task you need to remember.
What it displaces: A next-week priority with nothing displaced is a wish. Naming the cost is what makes it a plan.
Do not congratulate me: Assistants are tuned to be encouraging, and encouragement is noise in a review. You want the stall count, not a well done.

Before and after

The lazy version

Summarise what I did this week and suggest what to focus on next week.

What goes wrong: A warm paragraph about a productive week and three generic suggestions. Nothing is comparable to last week, so a task stalled for a month never once shows up as stalled.

The pattern, filled in

Worked example

Here is my raw material from this week: [my task list with strikethroughs, 9 git commits, two lesson-completion notes, a maths test score of 68, and three half-written notes.]

My current goals: (1) finish Phase 2 of the curriculum by 30 April — I would know by lesson-completion count; (2) get maths above 80 — I would know from test scores; (3) ship the classifier project — I would know when it runs end to end on new data.
Last week's digest, for comparison: [pasted]

Produce the digest in exactly this shape, every week:
1. SHIPPED  2. MOVED  3. STALLED (with week count)  4. AGAINST GOALS (with evidence)  5. PATTERN (vs last week)  6. ONE THING (and what it displaces)

Rules: do not congratulate me. Do not pad. If a section is empty, write "nothing".

What that buys you: The pattern line names it plainly: the classifier project has been in MOVED for three weeks running and has never once appeared in SHIPPED. That is the finding, and no amount of week-by-week enthusiasm would have surfaced it.

Now do it on your own work

Run this for three consecutive weeks, feeding each week's digest into the next. Do not change the format between runs. At week three, read all three digests together and see what only becomes visible across them.

Check yourself — every box should be true

You ran it three weeks in a row without changing the six sections.
You fed the previous digest in each time.
At least one item accumulated a stall count across the weeks.
You can name one thing visible across three digests that no single week showed.

Then explain it back

Something only became visible once you had three digests side by side. What was it, and why could a single week's review never have shown it?

Write two or three sentences in your own words. If you cannot, you copied a prompt rather than learning a pattern — which is the whole difference this library is trying to make.

Does this depend on which AI tool I use?

Any assistant, but continuity is the practical constraint: pasting last week's digest in yourself is more reliable than trusting a tool's memory feature, and it keeps the whole history in a file you own rather than inside one vendor. That portability matters more than any per-tool convenience here.

Why it actually works

Same review, same six questions, every week — small compounding effort beating occasional intensity. It is the same argument behind starting the curriculum early rather than harder.

Why small and regular wins →

Prompting well is a skill you can pick up in a week. Understanding what is happening underneath is the 18-month curriculum — and it is what stays useful when the tools change.

FAQ

Is this not just journalling?: Journalling is unstructured by design and is not meant to be compared. This is fixed-format specifically so that week five can be read against week one.
What if I had a bad week with nothing shipped?: Then SHIPPED says “nothing”, which is the most useful digest you will get. Two of those in a row is a signal, and the format is built so you cannot hide it from yourself.

Next patterns

The Prompt Patterns Library is a separate, optional track — it does not change the 78-week curriculum or your roadmap. See what the curriculum actually covers →