Plan & compute — how the deep dives fit together

Built 2026-10-20 from the seven deep-dive tabs. Estimates are labelled; the evidence and sources live in each tab.

Bottom line

• Verdict first (My take): no idea credibly reaches $10k a month. Only I1 and I2 are worth a cheap 4-week test; I12 and I10 are side bets; skip I5, I6 and I7. The plan below covers everything in case you want to run more than the tests.
• If you run tests, this order: I1 Apify ad-library actors → I2 Prozorro MVP → optionally I12 Diagram Blocks (it also tests whether Atlassian pays out to Ukraine) and I10 ProofCut. I5, I6 and I7 stay parked.
• Three shared engines carry seven ideas. The ad-library engine feeds I1, I5 and I10. The video engine feeds I6, I7 and I10. The Ukrainian document engine feeds I2 and later I4. Build each engine once.
• Your Claude Code subscription is the factory, not the engine. It builds, writes, publishes and maintains. The products themselves run with no model (scrapers, rules, renders, plugins) or on the customer's side (their agent, their key, their Rovo credits). Only three spots would need a paid model, and each needs your yes first (see D1–D2).
• The honest money picture (estimates from the tabs): I1 $2.5–5k a month in 6–9 months, I12 $1–4k after 12 months, I10 $0.5–4k MRR in year one. Their low ends add up to only about $4k. Reaching $10k needs I2 to convert (about 330–485 paying suppliers, 0.8–1.2% of competitive bidders; the tab gives no timeline) or the others to reach their upper ranges. I5, I6 and I7 are upside bets, not the plan.
• What only you can do: about one working day of one-time setup (accounts, ID checks, payouts), six decisions below, and a weekly 15-minute go/kill review.

What changed after the deep dives

Three engines, seven ideas

flowchart LR
  A["Ad-library engine: official APIs and actors"] --> I1["I1 Apify actors"]
  A --> I5["I5 free rankings, kits, community"]
  A --> I10["I10 ProofCut: what is scaling"]
  V["Video engine: renders, QA, style files"] --> I6["I6 video-dna and shorts-qa"]
  V --> I7["I7 character episodes"]
  V --> I10
  D["Ukrainian document engine"] --> I2["I2 Prozorro co-pilot"]
  D --> I4["I4 grant co-pilot, later"]
  P["Platform plugins, no model"] --> I12["I12 Atlassian apps"]

• Ad-library engine: one set of collectors (TikTok ads, Google Ads Transparency, LinkedIn Ad Library, Meta's official API) sells raw runs on Apify (I1). It produces the free rankings and teardowns (I5) and feeds ProofCut's "formats scaling in your category" (I10).
• Video engine: the beat-sync renderer, caption and QA checks and style files become open-source tools (I6). They also produce the character episodes (I7) and render merchants' ads (I10).
• Ukrainian document engine: tender ingestion, rules and registry checks, and local OCR for I2. The same parts later run the grant co-pilot (I4).

Where the AI compute runs

flowchart TD
  S["Your Claude Code subscription"] --> B["Build, write, publish, maintain"]
  B --> N["Products that need no model: I1 scrapers, I2 checks, I10 renders, I12 plugins, I6 shorts-qa"]
  B --> C["Customer-side reasoning: their agent, their key, their Rovo credits"]
  B --> Q{"Paid model at runtime?"}
  Q -->|"only with your yes"| Y["I2 document reading, I10 AI motion, I1 paid mode"]
  Q -->|"default"| X["Off"]

Why this matters: under your billing rule, Claude runs only on your subscription and only as the operator. Designing products so the customer's own agent does the thinking also keeps margins near 100% and removes the per-customer model cost entirely. That is the main reason the Apify, Atlassian and open-source plays lean towards agents.

The next 30 days

gantt
  title First 30 days from 21 Oct 2026
  dateFormat YYYY-MM-DD
  section You, once
  Accounts and ID checks       :s1, 2026-10-21, 3d
  section I1 Apify
  Build 4 actors               :a1, 2026-10-22, 10d
  Publish and measure          :a2, after a1, 18d
  section I2 Prozorro
  Ingest tenders and lookup    :b1, 2026-10-22, 14d
  Alerts and prepaid test      :b2, after b1, 14d
  section I12 Atlassian
  Vendor check and payout      :c0, 2026-10-21, 5d
  Build Diagram Blocks         :c1, 2026-10-24, 8d
  Marketplace review           :c2, after c1, 15d
  section I10 ProofCut
  Build on I1 ad data          :d1, 2026-11-01, 18d
  section Content
  I5 weekly issue starts       :e1, 2026-10-29, 23d
  I6 two repos                 :f1, 2026-11-04, 10d
  I7 two pilots                :g1, 2026-10-28, 33d
  section Decisions
  I5 gate                      :milestone, m1, 2026-11-20, 0d
  I1 and I2 gates              :milestone, m2, 2026-11-19, 0d
  I7 gate                      :milestone, m3, 2026-11-30, 0d

Go/kill gates (from the tabs):

• I1, day 28: keep any actor with 10+ outside paying users or growing 20%+ a week.
• I2, week 4: 10+ suppliers prepay 490–1,490 UAH, otherwise rethink the offer.
• I12: vendor payout to a Ukrainian account confirmed before app #2. Then 30 and 60 days after listing, check installs and trials against the tab's thresholds.
• I10, 28 days after approval: 40+ installs, 4+ paying merchants, and 25% of installers making 3+ ads.
• I5, 20 Nov: kits have earned ≥$500, 300+ free subscribers, and 15+ have prepaid for the community.
• I7, 30 Nov: pilot retention and comments clear the tab's thresholds, otherwise stop.

Your one-time setup

1. FOP group 3 with a USD account, plus Payoneer (all ideas).
2. Apify: billing details, ID photo, wire payout to the FOP account (I1).
3. Atlassian vendor account: company-domain email, Stripe Identity check (an ID check only, not a Stripe account), bank details. Confirm that a Ukrainian account works (I12).
4. Meta developer account and Ad Library API identity check. This decides whether I5 gets legal EU reach data.
5. monobank internet acquiring for UAH subscriptions (I2).
6. Shopify Partner account and payout method (I10).
7. Whop seller account with UAH payout (I5).
8. YouTube channel: phone verification, Cloud project set to "In Production" (I7, I5 video).
9. Own Spine licence if I7 uses rigs in code ($69 or $379).

Decisions only you can make

Sources

1. I1 · Apify actors — deep dive
2. I2 · Prozorro tender co-pilot — deep dive
3. I5 · Ad-intelligence data asset — deep dive
4. I6 · Open-source video pipeline + hosted tier — deep dive
5. I7 · Owned 2D character universe — deep dive
6. I10 · Product media for merchants, better variations — deep dive
7. I12 · Atlassian Forge apps — deep dive
8. Full report (draft v2)
9. My take: one-page verdict