Where we are
Session 3 fixed the order without changing a word. Tonight the words change, and for the first time in this course the improvement has a number on it rather than an opinion.
★ What you walk out with tonight. Your own message scored, the list of abstractions in it by name, a rewritten version, and a sentence you can use when somebody asks you to add more context.
LiveWhat is measured and what is a rubric2 min▶
Worth being precise about, because it is what makes the number worth anything:
- Genuinely measured, on your text: words per sentence, total words, how many dates and numbers appear, which corporate abstractions appear (from a fixed list of 30, reported back by name), and which sentence contains your ask, out of how many.
- Measured, but heuristically: the passive-voice count is a regex over a fixed participle list. It will miss some passives and occasionally flag something that is not one. Treat it as a prompt to look, not a verdict.
- A rubric, chosen not discovered: turning those six numbers into a score out of 100. The weights are written down in the source map. Two people can write a 70 and an 85 and the 70 can be the better message for its reader.
Nothing you paste leaves the browser. There is no model call anywhere in this widget - it is string analysis, which is also why it can be honest about its own limits.
Six things a reader is reacting to, whether or not they could name them
7 min liveLivePassive voice, and when it is actually evasion4 min▶
Passive voice is not wrong. "The database was migrated overnight" is fine - nobody cares who did it. It becomes a problem in one specific situation, and that situation is exactly the one Daybreak's data lead is in.
- "It was subsequently determined that the migration is more involved." Determined by whom? The sentence removes the person, and on a piece of bad news that reads as distancing yourself from it - whether or not that was the intention.
- "We underestimated the migration." Same fact, four words shorter, and it reads as someone in control of their own project rather than someone reporting weather.
- The rule of thumb: if the sentence is about something going wrong, name who. Passive on bad news is read as evasion by almost everybody, and it is the most expensive misreading in this session.
Daybreak's original has eight passive constructions in 163 words, and every one of them is on the part that went wrong. Nobody wrote that deliberately. It is just what people reach for when they are uncomfortable, which is precisely why it is worth counting rather than trusting your ear.
LiveThe abstractions, listed by name4 min▶
The widget checks against a fixed list of 30 phrases. It is a chosen list, not a derived one - a starting point for noticing rather than a standard. Here is what it finds in Daybreak's original, all nine of them:
circle back · going forward · holistic · ecosystem · workstream · state of play · knock-on · as you will be aware · at this moment in time
Notice the common property: each one could be deleted without losing information. "Circle back" means "write to you". "State of play" means "status". "As you will be aware" means "you know this already", which is a reason to delete the sentence rather than to introduce it.
Some of these are fine in some rooms. "Workstream" is a real word in organisations that have workstreams. The test is not the word, it is whether the reader learns anything from it - and the counter exists to make you look, not to make the decision.
That is one sentence from Daybreak's original: 33 words, three passive constructions, one abstraction, and no person anywhere in it. The rewrite is "We underestimated the migration. The payments integration took three weeks longer than we scoped." Thirteen words, two sentences, and somebody is responsible.
The clarity scorer
★ 22 min · everyone buildsTwo modes. The Daybreak message has levers that genuinely rewrite the text. Your own message does not, because nothing here can honestly rewrite something it has never seen - but every number is computed by exactly the same functions.
5 / 100 · ask in sentence 4 of 4 · 40.8 words per sentence · 8 passives · 9 abstractions · 0 specifics · 163 words. The message as sent, three times, to no reply. Nothing about it is unprofessional, which is exactly why it survived three attempts.
35 / 100 · ask now in sentence 1 of 4. One paragraph moved. Not a single word rewritten - the length, the passives and the abstractions are all still exactly as they were. Thirty points for a cut-and-paste.
45 / 100 · 117 words, 4 abstractions. Delete the opening that recaps what she already knows. Forty-six words gone and five of the nine abstractions went with them, because that paragraph was where they lived.
59 / 100 · 18.3 words per sentence. One idea per sentence. Watch the ask position: it stays at sentence 1 because you moved it first. Doing this step before session 3 would have taken it to sentence 9 of 9.
69 / 100 · 3 passives left. Say who did what. "We underestimated" instead of "it was determined". The message stops sounding like weather and starts sounding like a person running a project.
84 / 100 · 1 abstraction, 69 words. Plain words. The biggest single drop in length in the whole ladder, because abstractions are long as well as empty.
100 / 100 · 9.5 words per sentence · 0 passives · 0 abstractions · 7 specifics · 57 words. Dates and numbers instead of adjectives. 57 words, from 163, and it now says more than the original did.
Start with the message as sent and read the score. Five out of 100. Then read the message itself - it is not badly written by any normal standard. That gap between "reads fine" and "scores 5" is the reason this widget exists.
Switch on "put the ask in the first sentence" alone. 5 to 35. Nothing else on the board changes - same words, same length, same passives. Thirty points, one cut-and-paste, and this is the single highest-value thirty seconds in the course.
Now the counter-intuitive one. Reset, and switch on "one idea per sentence" alone. Score 19, and look at the ask line: sentence 9 of 9. Good editing, applied without fixing structure first, buried the ask further. This is why session 3 came before session 4.
Climb the whole ladder in order and watch the abstraction list shrink. It names them, so you can see exactly which phrase each rung removed.
Get to 100, then read the finished message out loud. 57 words. It contains more information than the 163-word version, which is the part people find hardest to believe until they see both.
Now switch on "add background so they have the full picture". 100 drops to 64. The ask goes back to sentence 4 of 9 and the message returns to 145 words. Read the paragraph it adds: every word is true, relevant and well written. It is in front of the ask, and that is all it takes.
Practise the sentence. Not "I think that is too long". Instead: "Happy to include it - can it go below the ask? Right now it puts the decision in sentence 4 of 9." You have accepted the request and protected the message, and nobody has to lose an argument.
Switch to "Paste your own" and put your session 3 message in. Read the six rows. The abstraction list is usually the surprising one, because it names phrases you did not notice yourself typing.
If you take one thing from tonight. "Your ask is in sentence 4 of 9" is a fact and it ends a discussion that "this feels long" would have kept alive for a week. Measure the thing, then make the request specific.
Before session 5
25 min- Score three real messages you have sent. Write the six numbers for each. Look for the pattern - most people have one row that is consistently bad and it is usually not the one they expected.
- Rewrite one to 80 or better and send the rewrite if it is still live.
- Collect your own abstractions. The widget's list is 30 entries and yours has different ones on it. Note the phrases you personally reach for.
- Find one passive on bad news in something you wrote, and rewrite it with a person in it. Notice how different it feels to send.
- Bring the 100-point version to session 5, because session 5 asks you to say it out loud, and a written message that scores 100 is not automatically a spoken one that works.
Three questions
Clear all three and session 4 gets stamped in your passport.
1 · One lever takes the message from 5 to 35 without changing a single word. Which, and why is it worth 30 points?
The words, the length, the passives and the abstractions are all identical before and after. Only the order changed. Ask position carries 30 points because it is the difference between a message that can be acted on and one that has to be finished first.
2 · Switching on "one idea per sentence" alone scores 19 and moves the ask to sentence 9 of 9. What does that teach?
Splitting is good editing. Position is measured relative to the whole message, so every sentence you add above the ask buries it further. Fix the order first and then the same edit costs you nothing - which is exactly the order these two sessions come in.
3 · Adding a paragraph of accurate, relevant background takes the finished message from 100 to 64. Why?
There is no padding detector and no penalty for thoroughness. The paragraph is honest and well written. It simply sits above the decision, and the measurement notices. The fix is not to delete it - it is to put it below the ask.
What this session teaches, and where it came from
Full rubric, every constant, and the honest limits of the passive and abstraction detectors: the course's source map.
Session 4 cheat sheet · pin this
You have a message that scores 100 on paper. Session 5 makes you say it out loud, where none of these measurements exist: the opening line, what has to be cut because a room cannot re-read, why the interruption is usually the real question, and using silence on purpose after you name a number.