oll.am · The single living plan · Flagship-Real Roadmap

oll.in, in order.

The one to-do list until a stranger completes the value ladder. oll.in is THE flagship — the agentic Swiss job-application agent for everyday job seekers. The whole platform it needs is already live. What's missing is the sellable front door. Three steps, in order.
Jump: The thesis The value ladder Step 1 · foto Step 2 · scout Step 3 · distribution Ship, not build

The flagship, in one paragraph

oll.in is the agentic Swiss job-application / career agent for everyday job seekers — the nurse, the teacher, the tradesperson, the admin — not for software engineers.

The whole platform it needs is already LIVE and proven: the three spines — Core (identity · billing · email, core.oll.am), Model (any model, one seam, model.oll.am), Memory (per-user corpus + cited retrieval, memory.oll.am) — plus the agent seam (OpenCLAW → oll-mcp → the products, proven end-to-end over stdio + HTTP). oll.in doesn't need a new backend. It needs a sellable front door.

It becomes "real" when a stranger completes the value ladder. Three steps, strictly in order — each one earns the right to build the next.

The value ladder — three rungs to "real"

CHF 29
Bewerbungsfoto on-ramp
The first franc. A professional application photo, pay-once. Lowest-friction way a stranger tries oll.in.
→
CHF 99
Application pass
The agent scouts + matches + prepares a full application. This is the flagship value.
→
CHF 19/mo
Career Radar
Ongoing: the agent watches the market for you. First recurring MRR.

Steps map one-to-one onto the rungs: Step 1 unlocks the CHF 29 on-ramp, Step 2 builds the scout half that powers the CHF 99 pass, Step 3 is the distribution that turns the pass into the CHF 19/mo Radar (recurring). Do them in order — do not skip ahead.

1 Un-park foto — the first franc 🔨 nearest dollar · CHF 29 Bewerbungsfoto on-ramp

Status — the backend is DONE, code-verified

foto-service is architecturally COMPLETE and code-verified. The mock provider runs the whole flow at zero spend; the REAL Replicate Flux-LoRA pipeline is pinned verbatim from wardrobai (ostris/flux-dev-lora-trainer, a per-order private destination model, trainer + sampler params byte-for-byte); foto owns its own results-ready email via Resend (Core's /email/send is recipient-scoped, so it can't email a guest); guest fulfillment is idempotent on job_id+session_id and amount/product-bound (a cheaper session can't fulfil a pricier tier). It is BLOCKED ONLY at the sell/infra gate — not on code.

Gap-closing checklist — in order, two lanes

🤖 Claude — code, this repo
  1. Re-land the Core guest-checkout pair. POST /api/billing/guest-checkout-session + GET /api/billing/guest-verify-session. They already exist on origin/feat/core-guest-checkout (commits c97e244 + d327745) but NOT on main (old PR #21 was closed/conflicting). Open a fresh clean PR off origin/main → green CI → adversarial verify (this is Core = the money path) → Sam merges (per-PR go required).
  2. foto-service Coolify deploy config. Base Directory /services/foto, its own Dockerfile, Watch Paths services/foto/**, the caller-side env matrix (CORE_BASE_URL=https://core.oll.am, FOTO_PUBLIC_BASE_URL, FOTO_PROVIDER, EMAIL_PROVIDER) + .env.example truth-up.
  3. Buyer-facing checkout page. Collects email + selfies + style, calls POST /api/foto/checkout, redirects to the Stripe URL, returns to /fulfill. This is the ONLY genuinely net-new code in Step 1 — foto is otherwise pure backend with no front door.
👤 Sam — console, ~15 min (outward/irreversible)
  1. Stripe prices. Create the CHF 29 (+49/79) ONE-TIME prices → set STRIPE_PRICE_HEADSHOT_CHF29_ONE_TIME (etc.) on Core; rotate STRIPE_SECRET_KEY test→sk_live.
  2. Live webhook. Register https://core.oll.am/api/billing/webhook (checkout.session.completed) → set STRIPE_WEBHOOK_SECRET on Core. Without it, payment succeeds but nothing confirms server-side.
  3. Replicate token. Drop REPLICATE_API_TOKEN on foto so real headshots generate. Without it foto safely serves mock images — fine for Claude's dev-test, NOT sellable.
  4. Coolify app + domain. Create the foto-service app from the deploy config; point shot.oll.am (instant — DNS already wildcards to the VPS, auto-TLS) → deploy.
✅ Proof of done — the definition of the first franc: real card → real headshot → delivered email.
2 Build the agent's job-SCOUT half powers the CHF 99 application pass

The agent has two halves. The drafting half is already LIVE — the oll-mcp platform verbs (clarify · rewrite · draft · spec · account), proven end-to-end through OpenCLAW. What's missing is the scout half.

Add legal job-search APIs — Adzuna / Arbeitnow / Jooble — so the agent can find + match roles to the user. Combined with the live drafting half, this is what powers the CHF 99 application pass: scout a role → match it to the user's memory corpus → prepare the application → the human sends it.

This plus the Step-1 checkout page are the only two genuinely net-new code pieces in the whole roadmap. Everything else is wiring already-live infrastructure.

3 Distribution playbook → CHF 19/mo Career Radar · first recurring MRR

This is Sam's documented bottleneck. The engineering is done; showing up is the gap. The plays:

  • Comparison pages — vs LinkedIn Premium · vs ChatGPT · AI Bewerbungsfoto vs the CHF 300 studio.
  • Free tools — lightweight, shareable, SEO-earning entry points.
  • The self-select live demo — let a visitor pick their profession and watch the agent work.
  • Swiss channels — German LinkedIn, profession communities, career-service partners.

Then the ladder's top rung lands: CHF 19/mo Career Radar = first recurring MRR — the moment oll.in stops being a one-off sale and becomes a subscription.

This is shipping, not building

foto's backend is finished. The entire roadmap above is the last mile of already-live infrastructure — not a new build.

The honest ship-vs-build read: the three spines + the agent seam are deployed and proven. The only net-new code in this whole plan is (1) the Step-1 buyer checkout page (required to sell) and (2) the Step-2 job-search APIs. Everything else is Stripe console moves, Coolify config, and pointing a domain. The risk here is not "can we build it" — it's the standing pattern of starting something new instead of charging for the finished thing. Don't. Ship Step 1.

Related, not duplicated: the per-product migration briefs, branch-consolidation order, and the multi-product-on-Core detail live in the Overnight Plan. The flagship narrative + distribution research is in the oll.in launch kit; the agent architecture is in The Agent Seam. Live status: Control Room.