Change the model
Move the sliders and see whether the business works. The presets load the three sizes the model was built for. The link in your address bar reproduces the exact scenario.
Chapter 5 · Peoplarium
Business model
Does it make money, and when?
With AI coding agents doing the typing, the MVP needs about 420 founder hours and about CAD 20k in cash, and launches around 5 January 2027 at 30 hours a week. Once Plus is live it breaks even near 4,600 monthly active members, and the share of members who pay moves the result far more than the price does.
As of September 2026 · Peoplarium is a Digi117 venture in development.
Net result a month, before founder pay+CAD 1,963
- Break-even
- 4,640 members
- Launch (agent fleet)
- January 5, 2027
- Founder hours to launch
- 396 h · 13.2 weeks
- Cost per member
- CAD 0.50
Launch is held at the floor: counsel, shadow moderation, the private beta and a security test take about 10 weeks, whatever the hours.
Rounded model from the Peoplarium estimate of 27 September 2026. Steady state with Plus live, before founder pay and marketing. Published ranges, not a forecast.
The model is rounded and uses published ranges. It is not a forecast. It assumes Plus is live, and it leaves out founder pay, marketing and taxes collected on subscriptions.
The estimate
The first estimate assumed a solo developer writing the code with AI help: about 1,120 founder hours and a public launch in late July 2027. The second assumes a fleet of AI coding agents writes and tests the code while the founder writes specs, reviews pull requests, tests on real phones and decides. That cut the hours from about 1,120 to about 420, with about 400 of them before launch.
| Phase | What it covers | Agent fleet, expected hours | Solo, expected hours |
|---|---|---|---|
| 0 Foundation | Counsel coordination, vendor setup, policy packs, UX and design system, privacy | about 70 | about 125 |
| 1 MVP | Everything in the web app, QA, shadow moderation, private beta | about 275 | about 780 |
| 1b Voice | Voice speed-meets and their moderation | about 20 | about 70 |
| Total with 15% contingency | about 420 | about 1,120 |
Expected hours use a three-point estimate: (low + 4 × likely + high) ÷ 6. The modules that shrink least are the ones where judgment is the work: policy packs, moderation, QA and the beta.
The timeline
Agents shrink the typing, not the waits. Four things set the floor: counsel turnaround of three to four weeks, at least two to four weeks of shadow moderation on real traffic, a private beta of about four weeks, and an external security test. Even at 40 hours a week the launch cannot come much before mid-December.
| Founder hours a week | Weeks to launch | Launch around |
|---|---|---|
| 20 | 20 | late February 2027 |
| 25 | 16 | late January 2027 |
| 30 | 13 | early January 2027 |
| 40 | 10 | mid-December 2026, near the floor |
Early January also happens to be the busiest stretch of the year for dating apps.
Cash before launch
About CAD 20k gets Peoplarium to launch. Digi117 already has the company, the bank account and the books, so incorporation drops out. The line worth defending is the external security test: agent-written code will hold verification results, consent records and moderation decisions.
| What | Share of the launch cash |
|---|---|
| External security test of the agent-written code | about a quarter |
| AI coding agents during the build | about a fifth |
| Counsel: terms, privacy policy, badge wording, extra review | about a quarter |
| Insurance, trademark, hosting during the build, test usage | about a seventh |
| Contingency, 15% | the rest |
Founder time is not in this number. It is about 420 hours.
Monthly result at three sizes
The business loses a little at 1,000 members and earns at 10,000 and 50,000, assuming 3%, 4% and 5% of members pay for Plus. AI model usage is the largest cost at every size, so model choice and caching matter more than hosting.
| CAD a month, rounded to 10 | 1,000 members | 10,000 members | 50,000 members |
|---|---|---|---|
| Paying share | 3% | 4% | 5% |
| Net revenue | 530 | 7,010 | 43,800 |
| AI models | 210 | 2,120 | 10,580 |
| Identity checks | 0 | 700 | 3,260 |
| Moderation, voice, messages, email | 70 | 530 | 2,670 |
| Fixed costs (hosting, agents for upkeep, insurance, legal reserve) | 1,110 | 1,700 | 3,320 |
| Total operating cost | 1,390 | 5,050 | 19,820 |
| Net result | −CAD 870 | +CAD 1,960 | +CAD 23,980 |
| Cost per member | 1.39 | 0.50 | 0.40 |
Revenue assumes Plus at CAD 19.99 a month or CAD 191.90 a year, with 30% choosing the yearly plan, less card and billing fees and a 2% refund reserve.
Break-even and what moves it
With Plus live, Peoplarium covers its operating costs at roughly 4,600 monthly active members, using the rates of the 10,000-member scenario. Founder pay needs more on top.
The table shows the net monthly result at 10,000 members. The bold cell is the base case.
| Paying share ↓ · price → | CAD 14.99 | CAD 19.99 | CAD 24.99 | CAD 29.99 |
|---|---|---|---|---|
| 1% | −3,700 | −3,300 | −2,800 | −2,400 |
| 2% | −2,400 | −1,500 | −700 | +200 |
| 3% | −1,100 | +200 | +1,500 | +2,900 |
| 4% | +200 | +2,000 | +3,700 | +5,500 |
| 5% | +1,500 | +3,700 | +5,900 | +8,200 |
| 6% | +2,800 | +5,500 | +8,100 | +10,800 |
| 8% | +5,400 | +9,000 | +12,500 | +16,100 |
The best performer in the category converts about 9% of users to paid. Levers, in order of impact:
- Paying share. Each extra percentage point of paying members at 10,000 is worth about CAD 1,750 a month. Plus has to be clearly worth paying for.
- AI model cost. At about CAD 0.21 per member a month it is the biggest cost. Caching, smaller models for scam scans and translation, and batching can cut it by half or more.
- Identity checks. The vendor's free 500 checks a month cover a small launch. Above that it is about CAD 0.47 per new member, paid once.
- Fixed costs. About CAD 1,100 a month at 1,000 members, half of it the agent fleet kept on for fixes and upkeep. Nothing here needs staff.
Risks to these numbers
The estimate is more reliable than the revenue. Build hours can be tracked and corrected weekly; paying share and member counts are guesses until the beta.
| Risk | Effect | What to watch or do |
|---|---|---|
| Review becomes the bottleneck | Agents write code faster than one person can check it | Specs and acceptance tests first; line-by-line review of security-sensitive modules |
| Security holes in agent-written code | A leak of verification results or consent records is not recoverable | External test before launch; no ID images stored |
| Fewer founder hours | Every 5 hours a week less adds about 3 weeks | Log hours weekly; cut Phase 1b or admin polish first |
| Paying share below 3% | Break-even moves above 10,000 members | Test the value of Plus in the beta |
| AI prices or usage higher | The largest variable cost grows | Per-member token caps, caching, cheaper models for routine tasks |
| Slow density without hosts | Losses run longer | Accepted trade-off; online surfaces first |
| Marketing is not modeled | Growth may need paid spend | Budget it once organic growth is measured |
The phases after launch
Ask about this, or request more analysis
Your message goes to Vadym Chernega, founder of Digi117, not a bot. Reply within one business day.