Body composition, and the charter rule we wrote against it
Charter rule C6 banned body composition framing at every age. Then we specified a product built on exactly those numbers. Here is the whole argument.
Published
8
articles, all human-reviewed
JESS MOVEBlog
Notes
The arguments behind the product.
Engineering notes and product decisions, written up properly — including the ones that made the roadmap slower and the ones we got wrong the first time.
Clusters
12
7 of 55 supporting subjects covered
Words
5.4k
no article below 600 words
Audit floor
90
score required to publish, out of 100
Database constraints found four bugs the same afternoon
We moved the platform invariants into database constraints. Postgres rejected four writes the same afternoon that the services had been allowing.
Accessible interface design: six modes, not a font size
Accessible interface design as a settings toggle produces a worse product for everyone. Six modes change register, density, mechanics and guardrails together.
The nudge we did not send: notification timing as a defect
Law 2 treats a notification fired into a moment you cannot move as a defect. We count silence as a successful outcome, and the held rate is a headline metric.
Why the habit streak forgives, and what that cost us
A habit streak usually works by making people feel bad. We built Grace Tokens, Flare Mode and a Bereavement Hold instead, and week-six retention improved.
Seated movement, and why five variants or it does not ship
Seated movement and three other variants are authored independently, never degraded from the standing one. Five or it does not ship, and there is no override.
Workplace wellbeing data without an employer dashboard
If an HR director can see one named person stopped moving in March, workplace wellbeing data is a liability. The individual view is absent from the types.
Food photo analysis cannot tell you the energy figure
Food photo analysis returns a range, its evidence source and a confidence level, and never narrows it to look tidier. Twelve dimensions, and no health score.
Editorial pipeline
An agent drafts. A person publishes.
Those are two different verbs, and the system does not let them collapse into one. The status machine has no transition from draft to published — a draft goes to review, and review is a name, not a flag.
01
The thinnest topic cluster wins
Six clusters, each with a pillar page and a set of supporting subjects. The agent is commissioned against whichever cluster has the fewest published articles relative to its subjects — not against whatever ranked well last month.
02
The SEO agent writes a draft
A frontier-tier call through the AI Gateway, ceiling 18 ACU, 90-second timeout. It gets the cluster, the pillar path, the sibling subjects and the banned lexicon. It returns title, description, keywords, a Markdown body and internal links.
03
The audit scores it, deterministically
Title and description length, slug format, word count, heading structure, keyword density, internal linking, cluster validity — and the banned lexicon, which is a blocker rather than a warning. Same draft in, same score out, every time.
04
One repair attempt, not a loop
A failing draft goes back with its own findings attached, and the repair is kept only if it scores higher. An agent that cannot fix its blockers on the second pass will not on the fifth, and every pass spends real money.
05
A named person publishes, or nobody does
The status machine has no edge from draft to published. Publishing requires a reviewer name and re-runs the audit at that moment rather than trusting the score stored when the draft was written.
The rule the agent cannot negotiate
blockerEvery banned term in the brand lexicon — the weight-loss vocabulary, the guilt vocabulary, the streak-threat vocabulary — is a blocker in the audit, not a warning. A draft containing one cannot reach review however well it scores on everything else.
This matters specifically for search, because those phrases have real volume and an optimiser pointed at traffic will reach for them. The rule lives in the audit function rather than the prompt: a prompt is a request, a function is a gate.
What the audit measures
- title 30–62 chars
- description 110–158
- body ≥ 600 words
- ≥ 3 sections
- no H1 in body
- ≥ 4 internal links
- density ≤ 2.5%
- banned lexicon
- normalised slug
Every finding carries a fix as well as a complaint — that is asserted in the test suite, because a finding you cannot act on is noise.
Topic clusters
Cover a subject completely, then link it together.
That is what search visibility is made of — not repeating a phrase. Each cluster has one pillar page and the articles that link up to it, and the emptiest cluster is what the agent gets commissioned against next.
What micro-movement actually is
2/5Pillar: /micro-movement · informational intent
- how to be more active with a desk job
- how long a movement break needs to be
- exercise without changing clothes or going to a gym
- what to do when you have no time to exercise
- movement snacks versus one long workout
Weight management you actually control
1/5Pillar: /body-balance · commercial intent
- why does my weight go up and down every day
- waist measurement versus BMI
- what is a realistic rate of weight loss
- how to keep progress when life gets in the way
- what to track instead of the scales
Reading a meal from a photograph
1/5Pillar: /foodlens · informational intent
- how accurate are calorie counting apps
- how to eat more plants without changing your diet
- reading a food label without the maths
- how much salt is actually in my food
- swapping one thing instead of changing the whole meal
Strength and balance, without a gym
0/6Pillar: /micro-movement · informational intent
- exercises to stop falling
- how many times should I be able to stand up from a chair
- chair based exercises for older adults
- how to improve balance at home safely
- how much protein do you need after 65
- what a falls assessment actually involves
Moving with a condition, not around it
0/6Pillar: /body-balance · informational intent
- exercise with arthritis when it flares up
- what to eat on appetite suppressing medication
- staying strong while losing weight on medication
- low salt eating when you have kidney disease
- exercise and type 2 diabetes for beginners
- eating well with pancreatic insufficiency
Movement after 65, seated and supported
2/4Pillar: /micro-movement · informational intent
- staying independent at home as you get older
- how to get moving again after an illness
- exercise when you use a walking frame
- what to do if you are afraid of falling
A coach that knows when to say nothing
0/4Pillar: /mova · commercial intent
- do fitness apps actually work
- why reminder apps stop working after a week
- how to build a habit that survives a bad week
- what to do when you have broken your streak
What a watch adds, and what it cannot settle
0/4Pillar: /wearables · commercial intent
- do I need a fitness tracker to get fitter
- why is my step count different on every device
- how many steps a day actually matters
- using a fitness app without a smartwatch
Moving with other people
0/4Pillar: /challenges · commercial intent
- step challenges that do not just reward the fittest person
- how to get a family moving together
- exercising with a partner who is a different fitness level
- workplace wellbeing challenges that people actually finish
Workplace movement programmes that are not surveillance
1/4Pillar: /industries · commercial intent
- what can my employer see in a wellbeing app
- sedentary risk in hybrid teams
- measuring a wellbeing programme honestly
- wellbeing benefits staff will actually use
Movement for children without body talk
0/4Pillar: /for-children · informational intent
- fitness apps that are safe for children
- getting children moving without talking about weight
- screen breaks children will actually take
- what a parent can see in a child’s health app
What this platform refuses to do
0/4Pillar: /assurance · commercial intent
- are health apps safe with your data
- what happens to health data in a fitness app
- how to tell if a health app is trustworthy
- health apps that do not sell your data
Analytics
What we measure, and what we refuse to.
A blog on a health platform that quietly profiled its readers would be an odd thing to publish a privacy policy next to. The analytics are built to the same standard as the product.
No cookie
Nothing is written to the reader’s device. There is no consent banner because there is nothing to consent to.
No stored address
The connecting address and user agent are hashed on arrival with a salt that regenerates daily. The raw values are never held in a variable that outlives the request.
A view is not a read
A read is 25 seconds visible and 55 per cent scrolled. Both numbers are reported, separately, and the gap between them is the interesting part.
Visible time only
A background tab accrues nothing. A page left open over lunch is not a two-hour read, and counting it as one flatters us in exactly the wrong direction.
90-day retention
Events older than that are dropped rather than archived. Uniqueness holds within a day and deliberately stops holding across days.
Nothing to sell
There is no identifier in the event, so there is no audience to build, no profile to enrich and no third party to share one with.
The consequence is a unique-visitor count that is honest within a day and deliberately meaningless across days. We would rather have a number we can explain than a funnel we cannot.
Try the thing these are about.
Everything written here describes a product you can be using in two minutes. The writing is free either way.
Free to create. 150 AI credits a month for 2 months, then AI features need a plan — everything that is not AI carries on either way. Premium is £5.99 a month and cancels in one click. Under 18 a guardian confirms before the account opens.