The Kindness Physics Playbook
A paint-by-numbers guide to testing whether conscious kindness compounds — using two lateral wedges, AI-built tooling, and hard binary gates that can't lie to you. 14 days. $200. One answer.
Does one deliberate act of conscious kindness create more than one act in return? If kindness is a >1x multiplier, Rich has a movement. If it's a terminal event that gets consumed and dies, no app, website, counter, or AI will fix that. This playbook tests the physics before building the machine.
- Kindness is always a conscious act. Even a smile is a deliberate state change. The platform must make people MORE conscious, not less.
- Every act requires labor. Kindness is physics — an external force applied to change an object's trajectory. Zero-effort kindness isn't kindness.
- People are in survival mode. 90% of humanity is below the Maslow line where this matters. Design for the 10%, not the 90%.
- Lowest friction wins. If it takes more effort than their current momentum allows, they won't do it. Period.
- Attach to existing rituals. Don't create new habits. Piggyback on grace, driving, church, dinner. The habit already exists — ride it.
- Reflection multiplies action. Rich's father made the family share a highlight at dinner. Recording kindness doesn't just track it — it changes behavior.
- AI should prove its positive potential. This is the counter-argument to Patricia. Show that AI can power something genuinely beautiful.
Semi-retired fathers, 55–70, who attend a weekly faith gathering, already practice a daily gratitude ritual (grace, prayer, journaling), have adult children, and have said some version of "I want to do more meaningful work with my time" in the last 6 months.
They have the time (semi-retired). The money (won't balk at $10 acts). The ritual (grace = existing attachment point). The spiritual framework (kindness is a calling, not a nice-to-have). The relational depth (know people well enough for personal acts). The ego security (past needing credit).
Estimated count in Rich's direct network: 8–15 men. That's not a bug. That's the validation pool.
Set Up the Tracking Infrastructure
Before any kindness happens, build the invisible plumbing that captures signal. Rich should never have to manually log data — the system does it.
What Rich Does (Human Work)
- Write down 20 names of people who fit the ICP (or close to it) from his direct network
- For each name, write ONE specific kind act he could do for that person — something personal, not generic
- Rank the 20 by how well he knows them (top 10 become Wedge A, remaining become Wedge B candidates)
What AI Builds (Jason's Work)
1. TRACKING PAGE (single HTML, publish to NowPage)
- One-page form at a clean URL: ideas.asapai.net/kindness
- Fields: "What happened?" (text or voice-to-text)
"Would you do this for someone else?" (Yes / Maybe / No)
"Can we share your story anonymously?" (Yes / No)
- Auto-timestamps, auto-generates unique visit ID
- Stores to Supabase table: kindness_signals
2. VOICE CAPTURE LINE (Twilio)
- A phone number that answers with:
"Tell us about a kindness someone showed you. 60 seconds."
- Transcribes via Whisper API, stores to same Supabase table
- Texts the caller back: "Thank you. Can you tell us who
did this kind thing? Reply with their first name."
3. PHYSICAL CARDS (20x)
- Design a simple 3x2" card Rich can print at home
- Front: "You don't owe me anything."
- Back: QR code pointing to ideas.asapai.net/kindness
- Rich prints 20, puts them in envelopes
4. DASHBOARD (single HTML, publish to NowPage)
- ideas.asapai.net/kindness-dash
- Shows: total visits, total submissions, voice calls,
referral names collected, "would you do this" breakdown
- Auto-refreshes from Supabase
- Rich checks this ONE page. That's his entire analytics.
If Rich can't write 20 names and 20 personal acts — if he stalls here — the project isn't ready. The constraint isn't technology, it's clarity of who Rich wants to serve. Go back to brainstorming.
Rich Goes First. 10 Personal Acts. No Pitch.
Rich performs 10 deliberate, personal acts of kindness for 10 specific people in his network. Each act must require Rich to actually know the person — not generic door-holding. A handwritten note. Picking up their grandkid. Dropping off their favorite coffee with a card attached.
The Rules
- One act per day for 5 days, two people per day, morning and afternoon
- Attach the physical card (QR code to ideas.asapai.net/kindness) to every act
- Say nothing about the project. Don't explain. Don't pitch. Don't ask them to do anything. Hand them the card and walk away.
- Log each act in a simple text file: who, what, when. That's Rich's private journal. 30 seconds per entry.
What Rich is Testing
Does a deliberate, personal act of kindness + a physical card with a URL create any downstream action from the recipient? Does the energy transfer, or does it get consumed and die?
What AI Does During This Phase
- Monitor Supabase table for new submissions hourly - If a new submission arrives, send Rich a Telegram alert: "Signal detected. Someone visited the kindness page." (Don't say WHO yet -- let Rich check the dashboard) - Daily EOD summary to Rich via Telegram: "Day 3: 2 page visits, 1 submission, 0 voice calls. Cards distributed: 6/10." - NO outreach to recipients. NO follow-up automation. This phase is 100% organic signal measurement.
Count the QR scans and form submissions from Wedge A.
0 out of 10 scanned or submitted = The physical card mechanic is dead. Kindness was received and consumed. No energy transferred to the platform. This is valuable data. Proceed to Wedge B only.
1–2 out of 10 = Weak signal. The act resonated for some, but the card-to-URL conversion is low. Note WHO responded — what do they have in common?
3+ out of 10 = Real signal. The kindness-to-action pipeline works. Double down on Wedge A mechanics in Phase 3.
Anyone submits AND names someone THEY were kind to = Compounding confirmed. This is the kill shot. The entire thesis is validated.
Ask People to Report Kindness They RECEIVED
This runs in parallel with Wedge A starting on Day 3. Different people, different mechanic, different psychology. Instead of asking people to report what they DID, ask them to report what was done FOR them.
The Setup
- Rich texts or calls 15 people (different from the Wedge A list) with one message:
"Hey [name], I'm working on a small project about kindness and I need your help with something that'll take 60 seconds. Call this number and tell them about the kindest thing someone did for you recently. That's it. [PHONE NUMBER] Or if you'd rather type: [URL]"
What Rich is Testing
Three things simultaneously:
- Do people respond at all? (basic engagement signal)
- Do they use voice or text? (channel preference data)
- Do they name the person who was kind to them? (the viral loop trigger)
What AI Does During This Phase
- Twilio voice line transcribes all calls automatically
- AI categorizes each submission:
TYPE: what kind of act (service, gift, words, time, etc.)
DEPTH: surface (held a door) vs. deep (showed up during crisis)
NAMED: did they name the giver? Y/N
CHAIN: did they say "I want to report someone too"? Y/N
- If someone NAMES a giver, AI drafts a message for Rich to review:
"Hi [giver name], someone told us about something kind
you did. We're not going to say who, but we wanted you
to know: it mattered. Would you like to hear what they
said? [Y/N]"
Rich must MANUALLY approve sending this. No auto-send.
- Dashboard updates in real-time on ideas.asapai.net/kindness-dash
Count responses out of 15 outreach messages:
0–1 responses = The ask itself has too much friction, even from Rich's warm network. The "report received kindness" frame doesn't motivate action. Kill this mechanic.
2–4 responses = Moderate signal. Look at WHO responded. Are they all the same archetype? What channel did they use (voice vs. text)? This tells you the format, not just the demand.
5+ responses = Strong signal. People will talk about kindness they received. The question is whether they'll name the giver.
If 3+ people name the giver AND 1+ giver says "yes, I want to hear what they said" = Viral loop confirmed. You have a self-recruiting mechanic. This is the foundation of a real product.
Read the Data. Make the Call.
Rich and Jason sit down with the dashboard. No feelings. No stories. Just numbers.
The Decision Matrix
| Signal | Wedge A (Debt) | Wedge B (Inversion) | What It Means |
|---|---|---|---|
| Both dead | 0 scans | 0–1 replies | The thesis is wrong for this ICP. Kindness doesn't compound through Rich's network in this format. Pivot the ICP or the frame entirely. |
| A dead, B alive | 0 scans | 5+ replies | People will TALK about kindness but won't ACT on a card. Build a listening platform, not an action platform. Content/media play, not an app. |
| A alive, B dead | 3+ scans | 0–1 replies | Personal acts compound but people won't report on request. Build around the physical token/card mechanic. The coin idea has legs. |
| Both alive | 3+ scans | 5+ replies | Kindness compounds AND people will talk about it. Build the full platform. You have both the supply (acts) and the content (stories). |
| Compounding confirmed | Someone reports THEIR act | A named giver joins | The chain sustains without Rich pushing it. This is a movement, not a product. Go fast. |
What AI Builds Based on the Decision
WEEK 3 BUILD (only if Phase 3 shows signal):
1. KINDNESS STORY PAGE (NowPage, auto-published)
- AI takes approved submissions and generates
anonymous 30-second story cards
- Published to ideas.asapai.net/stories
- One new story per day = Rich's "daily kindness read"
2. WEEKLY BRIEFING (Forge auto-task)
- AI compiles weekly stats + best story + insights
- Sends to Rich via Telegram every Sunday
- Includes: "This week, X acts were reported.
Y people named someone. Z chains formed."
3. CHURCH PILOT LANDING PAGE
- Simple QR-code page for ONE church
- "What kind thing did you do or see this week?"
- Pre-service prompt, pastor script included
- Only build this AFTER organic signal confirms demand
4. PATRICIA DEMO PAGE
- A beautiful single page showing what AI built:
the tracking, the stories, the counter, the data
- Specifically designed for Rich to show Patricia
"This is what AI can do for kindness"
- Published to ideas.asapai.net/kindness-ai-demo
What This Costs
| Item | Cost | Who |
|---|---|---|
| Twilio phone number + transcription | ~$5/mo | Jason builds |
| NowPage hosting (existing) | $0 | Already running |
| Supabase tracking table | $0 (free tier) | Jason builds |
| 20 printed cards | ~$3 | Rich prints |
| 10 personal acts of kindness | ~$50–150 | Rich does |
| Claude API for transcription/categorization | ~$5 | Forge handles |
| Rich's time: 30 min/day x 14 days | 7 hours total | Rich |
| Jason's build time: Phase 0 infrastructure | ~4 hours | Jason |
| Total | ~$200 + 11 hours |
The Punchline
In 14 days, Rich will know one of four things:
- Kindness compounds AND people talk about it → Build the full platform. Rich has both supply and demand.
- Kindness compounds but people won't report → Build around physical tokens (the coin idea). The medium is physical, not digital.
- People talk but kindness doesn't compound → Build a content/media play. Podcast, newsletter, story page. Not an app.
- Neither works → The ICP or the frame is wrong. Pivot before building anything. This answer saves Rich months of building the wrong thing.
Every outcome is a win. Three of four give Rich a clear build path. The fourth saves him from building something nobody wants. There is no losing move in this playbook — only information.
The question was never "should Rich build a kindness app?" The question is: does kindness compound? Run the experiment. Let the physics answer.