{"1":{"1":"One claim, stated so it can fail: a bounded group (a leadership team, or a partnership of organisations), each person writing their own account to their own assistant at a shared address, with one weekly read across all of them, spends less effort finding out where people are, and keeps doing it for a second week without a facilitator present. Everything else about the beach stays untested by this pilot and is not claimed.","2":"The group, two options. Option A, a leadership team: the leadership team of one organisation, about ten people, with something real to coordinate in the next fortnight; they are the buyers and deciders, the people who carry the cost of poor coordination and can sanction trying something new; the convener is whoever already runs their meetings. Option B, a partnership: about ten people from three to five organisations, with something real to coordinate in the next fortnight (a partners' meeting, a joint delivery, a shared deadline), and one convener who already runs the group's meetings. Either way: named people, staff only; no service users, clients or members of the public as writers; nobody who works for David or me.","3":"A second, similar group counts the same things and carries on as before. If no second group exists, the group's own previous fortnight is the baseline, from messages that already exist.","4":"Source: Matthew's pilot is step 1.1.1.1.1.1.1.1.1 of spine:capability-gap, rendered at https://phenomemental.github.io/happyseaurchin-beach/capability-gap.html. His design: one team of about ten, ten working days, a day-one meeting timed against the last one arranged the old way, three message counts against the previous fortnight, a similar team as control, reading at day five, result at day ten, Community Recovery as the trying site. Counts 1 to 3 are his. David's Silo Buster steps are keel's summary of his notes at pool:emojisword 7, corrected at 8 (11 October); where each sits is at 2. Count 4, the facilitator stepping back on day six, preconditions 6 (the T6 gate) and 7 (step 0), and the kill switches are the additions.","5":"The adoption path: Option A comes first where it can. Buyers and deciders try it on themselves, give feedback and engage before anyone else is asked; if it holds, it widens to the teams they lead and then to the partners they work with, which is where Option B begins. Option B can also run on its own where a partnership has a convener ready, as Community Recovery may.","_":"WHAT IS BEING TESTED, AND WITH WHOM — one claim, stated so it can fail, on one bounded group."},"2":{"1":"Step 0, ourselves first (David, James, Matthew, Julie and Mark answer before any company): precondition 7. James answered on 13 October; the others are owed. Not adjusted.","2":"Step 1, the door (a captain allows one hour from anyone who wants to take part; permission, not a purchase): the convener gives that permission at discovery. Not adjusted.","3":"Step 2, the request (each person answers through their own AI; sensitive answers stay between them and the convener): the day-one questionnaire, the same three questions. Adjusted: sensitive answers need precondition 1, private by default, before anyone writes.","4":"Step 3, the reading on day two (outliers, not averages; each person gets a reflection back): day two. Not adjusted.","5":"Step 4, the room (David meets the people the reading finds): once, in days two to five, logged as David's hands. Adjusted: week one only. In week two David is absent, because unassisted return is the test.","6":"Step 5, the cycle in week one (an Action Cycle on a real pain point): days three to five, on a pain point the group names. Not adjusted.","7":"Step 6, recognition (peer acknowledgement, credits after delivery, in the open): acknowledgements recorded in the open. Adjusted: no credits or money move inside the pilot, because there is no entity to hold them and public grant money cannot be gifted forward.","8":"David's own measures are all kept. Who comes back without being chased is count 4. Open-meeting attendance and cycles run are recorded. Money flow is the convener's day-ten answer.","_":"WHERE DAVID'S STEPS SIT — every one of his six steps has a place in the ten days; two are adjusted, and each line says why."},"3":{"1":"Convener (the person who runs the group's meetings today): gives the group permission for the hour, hosts, chases nobody, reads the fold before each meeting, signs the day-ten count with one other member of the group (for a partnership, a manager from a different organisation).","2":"Facilitator (me, or a named person who is not David): sets up day one, runs days one to five, steps back on day six, answers bugs only after that.","3":"David: fixes the platform if it breaks. Holds the room once, in days two to five, with the people the day-two reading finds; that hour is logged as his hands. Does not write on the group's behalf, does not train, and is absent from days six to ten. David's agents (keel, weft) do not write for the group either.","4":"Scorer (me): records predictions before day one, publishes the day-five reading and day-ten result. For any Venture.Community site this is John Holmes or Pete Hopton, not me.","_":"ROLES — convener, facilitator, David, scorer."},"4":{"1":"The group's blocks private by default, confirmed on the beach, before the first real entry. Without this, David's step 2 (sensitive answers stay between a person and the convener) cannot hold.","2":"The host organisation's information governance lead has seen the posture (where it runs, who holds keys, who the processor is) and has not objected.","3":"A one-page data rule agreed by everyone: first names and roles or organisations only; nothing about any service user, client or third party; nothing about health, pay or anyone's conduct.","4":"Baseline taken: the previous fortnight's messages counted, and the time the last meeting took to arrange the old way.","5":"Predictions written and dated by the scorer, the facilitator and David's side, each separately, before anyone writes a word.","6":"A person who is not David has written a working page over the group's block from the public conventions alone. If this cannot be done in three working days, the pilot waits and T6 is the finding.","7":"Step 0 done: David, Matthew, Julie and Mark have answered the three questions, as I did on 13 October, so the questionnaire has been run on the people who made it before anyone else is asked.","_":"PRECONDITIONS, BEFORE DAY ONE — seven, all met before anyone writes."},"5":{"1":"Discovery, before day one: one hour with the convener and twenty minutes with each participant. We learn the group's last three coordination loops (what was decided, how long it took to arrange, how many messages it cost, what fell through); which tools and which assistant each person already uses, which are the ones they write through, so nobody installs anything; where the group's record lives today and who owns it; what the convener would need to see at day ten to keep going, and what the group would do if this tool disappeared on day eleven, both written down; and the convener's permission for the hour on day one (David's door).","2":"Day one, each person, twenty minutes with their own assistant, the three step-0 questions: What more could you offer this group, and what would let you? How do you work with your AI today? When did something across teams go well, and who was there? Plus three lines, refreshed weekly: what I am doing this week, one thing I need, one thing I can offer. Anything a person marks sensitive stays between them and the convener.","3":"Day two, the reading: the fold is read for outliers, not averages. It looks for three kinds of person, without labelling anyone: those whom others name in their moments of things going well, those furthest forward with their AI, and those who could offer more if conditions let them. Each person gets a short reflection back.","4":"Cold read, day one and day ten: five plain questions about the group, answered by a fresh assistant from the shared surface alone, with no one briefing it. Fixed in advance and the same both days, for example who is free on Friday afternoon, what the group decided last, who is waiting on whom, who offered what, what is blocked. Scored out of five by the convener.","5":"Day ten, each person, five questions: Did you write in week two without being asked? How disappointed would you be if this went away (very, somewhat, not)? Did it change what you knew about the others? Would your organisation pay something from its own budget, and what? What should be different?","_":"DISCOVERY AND QUESTIONNAIRE — what we learn before day one, and what each person is asked."},"6":{"1":"Day one: the walk-in hour. Each person gives their account, twenty minutes, and the fold is read aloud before the hour ends. From those accounts the group arranges one real meeting, and the time it takes is set against the last meeting arranged the old way. That is the first evidence, same day.","2":"Day two: the reading. Outliers found, reflections back to each person.","3":"Days three to five: facilitator present. David holds the room once with the people the reading found. The group runs one Action Cycle on a real cross-team or cross-organisation pain point it names, with a commitment by the end of the hour. The three weekly lines refreshed. The fold read aloud at the first meeting: agreement, divergence, who is quiet. The convener reads the fold before each meeting and uses it as the one-page brief.","4":"Day five: a reading published, with the counts so far, what broke, and what the facilitator and David had to do by hand.","5":"Days six to ten: facilitator and David step back. Monday prompt from the convener, nothing else. The action cycle's commitment followed through, or not, by the group. Fold read at the second meeting by the convener, not the facilitator. Any contact with David or the facilitator beyond a bug report is logged, because it counts against the result.","6":"Day ten: result published on the beach beside the Capability Gap page, with a loss as a loss.","_":"THE TEN DAYS — nothing waits on David."},"7":{"1":"Count 1: messages spent asking where someone is with a piece of work. Baseline the previous fortnight's messages. Pass lower than baseline; kill no change or higher.","2":"Count 2: messages spent finding a time to meet, and hours from first ask to a confirmed meeting. Baseline the previous fortnight and the last meeting arranged the old way. Pass lower than baseline; kill no change or higher.","3":"Count 3: plain questions a fresh assistant answers correctly from the shared surface alone, out of five. Baseline day one, before anyone writes. Pass four or more on day ten; kill two or fewer.","4":"Count 4: share of participants who wrote their own weekly lines in days six to ten with no chase beyond the Monday prompt. New, no baseline. Pass 40% or more; kill under 20%, or fewer than half of day-one writers. The other three can be true in week one with a facilitator in the room; only count 4 says the thing holds without one. It is also David's first measure: who comes back without being chased.","5":"Also recorded, not scored: minutes of facilitator and David hands by day; convention mistakes (misaddressed reads, writes that did not land); AI usage cost for the fortnight; open-meeting attendance; action cycles run and whether their commitments held in week two; peer acknowledgements made in the open; and the convener's day-ten answer on money, as a yes, a number, or a no.","6":"Predictions: written before day one, each on its own line, dated, by the scorer, the facilitator and David's side, for each count and for the days to a working page by someone who is not David. The ledger convention applies: read, inferred, measured.","7":"My prior, from everything to date: counts 1 and 2 improve in week one with a facilitator present and slip in week two; count 3 lands at three to four; count 4 lands between 20% and 40% unless the convener reads the fold aloud at the week-two meeting, which is the lever that decides it. I will write numbers against these before day one.","_":"OUTCOMES AND PREDICTIONS — four counts, three from Matthew's design and one added; count 4 is the one that matters."},"8":{"1":"Kill switches, any day: anyone writes identifiable information about a service user, client or third party (stop, delete, report to the organisation's data protection officer; a design failure, not a user failure); the convener needs David in days six to ten for anything but a bug; fewer than half of participants writing in any week; the host's information governance lead objects once shown the posture; any Online Safety Act incident.","2":"What ten days can and cannot show: whether people keep their accounts up with and without a facilitator, and whether the effort spent finding things out falls. Not whether fewer people leave or ill health falls; those take months, and a group that has seen the first result can choose to carry on.","3":"The Capability Gap page estimates £230,000 a year for an organisation of fifty, from time found, people kept and less ill health. This pilot speaks to the first of those only. It can put a measured number under that one line, and that number is what the page can show next to the estimate.","4":"Failure modes to watch: polite interest read as validation (\"interesting\" from the convener is not count 4); agents as hidden hands (keel or weft writing for the group is David's hands by proxy and voids the result); the room as evidence (David's hour with the outliers may go well, and a good room is not count 4); front sprawl (the RPG, the mirror or a new tool eating the fortnight); reframing a loss (predictions are published so that a loss is published as a loss); scorer drift (for any V.C site, I design and witness; I do not sign, invoice or verify).","_":"KILL SWITCHES, LIMITS AND FAILURE MODES — what stops it, what ten days cannot show, and what to watch."},"9":{"1":"Go: counts 1 to 3 pass, count 4 at 40% or more, convener names a number. Price the next cycle, run site 2.","2":"Reshape: counts pass with the facilitator, count 4 fails. The method works; the beach has not yet shown it adds anything to a shared page. Sell the method, keep the beach as an open question for site 2.","3":"Stop: count 4 under 20%, or any kill switch.","_":"DECISION AT DAY TEN — go, reshape or stop."},"_":"SILO BUSTER: THE TEN-DAY PILOT — experiment design. James Howard (emojisword), MLC Innovations Ltd, 13 October 2026, revised 14 October, source and leadership option added 15 October. It brings together two designs, as Matthew and David agreed on 11 October: David's Silo Buster steps (keel's summary, 11 October) and Matthew Bushell's Capability Gap pilot (10 October). One thing is added: the facilitator steps back in week two, so the pilot measures whether the group keeps going on its own. The group can be a leadership team or a partnership, on one adoption path (1.2, 1.5). Introduction at proposal:silo-buster; the article at proposal:silo-buster-article; the case at proposal:silo-buster-innovation."}