Method: Mine actual revenue data from indie products. List every shared assumption from v1-v3. Invert each assumption. Generate ideas from the collision of proven revenue patterns and broken assumptions.
Date: 2026-04-09
We mined ~70 products with disclosed revenue from Gumroad, Etsy, IndieHackers, ProductHunt, and AppSumo. The patterns that matter:
| Job | Example | Revenue | Price |
|---|---|---|---|
| "Decide for me" | Canva social media templates | $1.4M in 3 years | $499/yr |
| "Make me look good" | iOS icon set (monochrome aesthetic) | $397K in 9 months | $28 |
| "Skip the learning curve" | English speaking pattern guide (PDF) | $568K total | $290 |
| "Give me the system" | Notion "Ultimate Brain" template | $2.1M in 2 years | $49-179 |
| "Time arbitrage" | Personalized coloring books | $839K total | $9.99 |
The highest revenue-per-complexity products are embarrassingly simple. Complexity is inversely correlated with revenue per hour invested. The products that work are not the ones that impress engineers — they're the ones that eliminate a specific moment of friction for a specific person.
The activation energy gap IS the product. Every successful product monetizes the distance between what people could do themselves and what they will actually do.
Every idea across all 3 brainstorms shares these assumptions. None were questioned.
| # | Assumption (v1-v3) | Inversion |
|---|---|---|
| 1 | The product should be a SaaS tool or platform | What if it's a digital product, template, plugin, or service? |
| 2 | The user is a creator, marketer, or knowledge worker | What if it's plumbers, parents, gamers-as-gamers, students? |
| 3 | AI should be the core value proposition | What if AI is invisible infrastructure, not the selling point? |
| 4 | The product needs to be novel/innovative | What if you clone a proven model into an underserved niche? |
| 5 | Revenue = subscription or transaction-based SaaS | What if it's a one-time purchase, ad revenue, or project fees? |
| 6 | Build one product | What if you build a portfolio of 5-10 micro-products? |
| 7 | Target a large market (>$500M TAM) | What if you own a $5M niche completely? |
| 8 | Scale before profiting | What if it's profitable from week 1? |
| 9 | Needs a growth loop or viral mechanic | What if distribution is someone else's problem (Etsy, AppSumo, YouTube SEO)? |
| 10 | Must reach $1M+ ARR | What if $50K MRR ($600K ARR) is the right target for 2 people? |
| 11 | Build everything from scratch | What if you build ON Etsy, Gumroad, YouTube, AppSumo? |
| 12 | Horizontal product | What if you go absurdly deep in one vertical? |
The biggest blind spot: v1-v3 all assumed the answer is a product. The revenue data suggests the answer might be a service, a media business, or a digital goods factory.
These are structurally different from every v1-v3 idea. None are "AI SaaS tools for creators."
What: Leo and Oskar run a productized AI agency offering 3-5 fixed-price services: - "We'll set up an AI chatbot for your business" — $1,000-2,000 - "We'll automate your content workflow" — $2,000-4,000 - "We'll build your MVP in 2 weeks" — $5,000-10,000
Oskar films each client engagement as YouTube content ("I Automated This Business in 48 Hours — Here's How"). Each project is simultaneously revenue AND marketing.
Why this is different from everything in v1-v3: It's not a product. It's a service. It generates revenue from day one. It builds domain expertise that reveals what product to build later. Most great SaaS companies started as agencies or consultancies — they productized their most-repeated service once they saw the pattern. HubSpot started as a marketing agency. Basecamp started as a web design firm.
The revenue math: 4 client projects/month at $3K average = $12K/month immediately. Oskar's YouTube content from the work drives inbound leads. As patterns emerge (which service do clients want most? which part is most templatable?), that's your product.
Assumptions inverted: #1 (not SaaS), #5 (project fees), #8 (profitable immediately), #4 (not novel — agencies exist; the AI acceleration is new)
The honest risk: Agencies don't scale. Trading time for money. But: you're trading time for money AND information AND content. The agency is the research method. You stop when you've found the product.
What: A plugin for professional video editing software (Premiere Pro, DaVinci Resolve) that automates the tedious parts: auto-remove silences, auto-generate captions + chapter markers, auto-suggest cuts based on audio energy, auto-color-match between clips. One-time purchase: $79-149. Annual updates: $39/yr.
Why this is different: It's a plugin INSIDE an existing workflow, not a standalone tool. The AI Photoshop script (Nano Banana) made $586K with this exact model. Editors don't want new tools — they want their current tools to be faster. The key insight from the revenue data: "Let me use AI without leaving my existing tool" is worth $50-150 as a one-time purchase.
Who: Video editors who use Premiere/DaVinci daily. Oskar IS this person. His audience IS these people.
Assumptions inverted: #1 (plugin, not SaaS), #3 (AI is invisible — it's "a better editing plugin," not "an AI tool"), #5 (one-time purchase), #9 (distribution through Oskar's YouTube + Adobe/Blackmagic marketplaces)
The honest risk: Premiere/DaVinci have APIs but they're limited. Building a reliable plugin requires deep understanding of their SDKs. Also: Adobe is adding AI features natively (Premiere's AI Assistant). The window may be 12-18 months.
What: A premium starter kit for building AI-powered web apps. Pre-built: authentication, Stripe billing with usage metering, AI chat interface, RAG pipeline, MCP server integration, deployment to Vercel/Railway, admin dashboard, email templates. One purchase: $199-349. The buyer saves 3-4 weeks of boilerplate work.
Why this is different: ShipFast (a Next.js boilerplate by Marc Lou) made $250K+ in revenue. It's a proven model. Leo can build a BETTER version specifically for AI SaaS — the hottest category. This isn't a tool; it's a product that ships once and sells forever.
Who: Developers building AI SaaS products. Oskar's tech audience. Every vibe-coder who outgrows Lovable/Bolt and wants a real codebase.
Assumptions inverted: #1 (digital product, not SaaS), #4 (not novel — deliberately cloning a proven model), #5 (one-time purchase), #8 (profitable from first sale)
The honest risk: Boilerplate market is competitive (ShipFast, LaunchFast, Shipixen, etc.). Differentiation is "specifically for AI SaaS" which is strong right now but may commoditize. Also: this is a developer tool, which means Oskar's audience overlap is partial.
What: Leo + Oskar build and operate a network of 5-10 faceless YouTube channels. Each channel covers a specific niche: tech explainers, crypto news roundups, gaming history, AI news, space facts, weird science. AI generates scripts (Claude) + voiceover (ElevenLabs, $11/mo) + visuals (stock + AI-generated). Oskar's YouTube expertise guides strategy, titles, thumbnails. Leo builds the automation pipeline.
Revenue: YouTube ads + sponsorships + affiliate links. Target: $5K-10K/month per channel once mature (3-6 months). Network of 5 channels = $25K-50K/month.
Why this is different: It's not a product OR a service. It's a media business. The "product" is the content. The "distribution" is YouTube's algorithm. The "moat" is: (a) Oskar's understanding of what YouTube promotes, and (b) Leo's automation pipeline that reduces production cost to near-zero.
Assumptions inverted: #1 (media business, not SaaS), #3 (AI is invisible infrastructure), #5 (ad revenue), #8 (revenue from month 1 of publishing), #11 (built on YouTube)
The honest risk: They killed "AI podcast video generator" before due to low margins. The difference: they're not selling a tool, they're running channels. COGS per video with AI: $1-5. Revenue per video with 50K views: $200-1,000. Margins are high. But: YouTube algorithm changes, content fatigue, and the "AI slop" crackdown could hurt. Also: channel growth takes months, which clashes with their "ship fast" DNA.
What: Leo builds an AI system that generates high-quality templates at scale: Notion templates, Google Sheets, Canva designs, Figma kits, Obsidian vaults. Instead of selling one template, they produce templates across 50+ niches (wedding planning, contractor project tracking, podcast launch kit, YouTube content calendar, crypto portfolio tracker, etc.) and sell across Etsy + Gumroad + their own store.
Revenue: $15-49 per template. High volume across many niches. Target: 50 templates generating $200-500/month each = $10K-25K/month.
Why this is different: Thomas Frank makes $120K/month selling Notion templates — but designs each one manually. This is the FACTORY model: AI generates the 80%, a human does the 20% (quality control, taste). The competitive advantage isn't any single template — it's the production system.
Assumptions inverted: #1 (digital products, not SaaS), #5 (one-time purchases), #6 (portfolio of many products, not one), #9 (Etsy/Gumroad handle distribution), #11 (built on existing platforms)
The honest risk: Template market is getting flooded with AI-generated crap. Quality control is the differentiator — and it's hard to maintain at scale. Also: Notion/Canva could build their own template generators, collapsing the market.
What: Pick ONE offline business vertical (dental offices, hair salons, yoga studios, auto repair shops — choose based on willingness to pay and pain intensity). Build a dead-simple AI assistant that handles their #1 admin headache: answering phones, booking appointments, sending reminders, handling cancellations. Not a platform. Not configurable. One vertical, perfectly tuned.
Revenue: $99-299/month per business. Target: 200 businesses = $20K-60K MRR.
Why this is different: CottageKeeper makes $4K MRR from 120 customers with cleaning checklists. DoggieDashboard makes $9K MRR for groomer bookings. The revenue data screams: tiny vertical tools for offline businesses print money because the incumbents (Salesforce, HubSpot) are too complex and too expensive.
Assumptions inverted: #2 (offline businesses, not tech workers), #4 (not novel — just well-executed for one niche), #7 (small niche), #12 (absurdly vertical)
The honest risk: Distribution. Oskar can't reach dental office managers through YouTube. This would require cold outreach, Google Ads, or partnerships. Mismatches the team's distribution advantage unless they find a vertical that overlaps with Oskar's audience (streamers? gaming cafes? content studios?).
The problem isn't the ideas. It's the frame. All three sessions tried to find THE IDEA — a single, novel, venture-scale product that would be "the next thing." This frame is wrong for a 2-person team that ships fast and has killed 5 products.
The revenue data shows a different path: most successful 2-person businesses didn't start by building a product. They started by doing something that made money, then noticed what worked, then productized THAT.
Ordered by "how fast does this make money and teach you what to build next":
| Rank | Idea | Time to first $1K | What you learn |
|---|---|---|---|
| 1 | AI Agency | 1-2 weeks | Which AI service people pay most for. That becomes your product. |
| 2 | EditPilot | 4-6 weeks (build + launch) | Whether editors pay for in-workflow AI. Proven by $586K Photoshop script. |
| 3 | ShipKit | 3-4 weeks (build + launch) | Whether devs pay for AI-specific boilerplate. Proven by ShipFast ($250K). |
| 4 | ClipFactory | 2-3 months (channel growth) | Whether AI-produced content monetizes at good margins. Media business, not product. |
| 5 | TemplateForge | 2-4 weeks (produce + list) | Whether AI-generated templates sell alongside human-made ones. |
| 6 | VerticalBot | 6-8 weeks (build + find customers) | What offline businesses pay for. Distribution is the bottleneck. |
Week 1-2: Run the agency. Oskar posts a YouTube video: "I'll Automate Your Business With AI — Free for the First 5 Clients" (or $500 for the first 5). Leo delivers. Film it all. Simultaneously:
Week 1-2: Launch ShipKit or EditPilot. Leo spends 2-3 weeks building whichever he feels more pull toward. This runs in parallel with the agency work and produces revenue from a different model (product vs. service).
Week 3-4: Read the signal. Which agency clients are most excited? What do they keep asking for? What are they willing to pay the most for? That's your product. Meanwhile, did ShipKit/EditPilot sell? What's the conversion rate?
Week 5+: Double down on what worked. Kill what didn't.
This approach generates revenue from week 1, produces content from week 1, and generates the demand signal that 3 brainstorm sessions couldn't produce — because the signal comes from real humans with real wallets, not from desk research.
v4 is deliberately short. The diagnosis: more ideas won't help. Getting money from real people will. The agency model is the fastest path to both revenue and product insight.