Build and Ship an AI SaaS in 48 Hours · 20 min · None yet — this lesson is pure product scoping
Objective
What you're building
A completed Product Definition Canvas — no code today. This course's reference product is an AI Proposal Generator (client name + project description in, a polished proposal document out); you can build that exact product or swap it for your own idea using the same architecture. Either way, today's output is a decision document, not a line of code.
The out-of-scope list matters as much as the feature list
5 features IN and 5 features OUT, written down before you start
Every scope-creep moment in a 48-hour build happens because "just one more feature" felt small in the moment. Writing your 5 out-of-scope items down NOW (before you're 6 hours in and tempted) gives you something concrete to point back to when a good idea shows up mid-build. For the proposal generator: IN is auth, 3 free proposals, paid unlimited tier, AI generation, PDF export. OUT is team accounts, proposal templates library, e-signature integration, CRM integration, white-labeling — all genuinely good ideas, all explicitly saved for later.
Why AI SaaS MVPs are different from traditional SaaS MVPs
No training data, no cold-start problem -- the AI is capable on day one
A traditional SaaS MVP (a scheduling tool, a CRM) needs real usage data before it's genuinely useful — the cold-start problem. An AI SaaS wrapping Claude is capable of good output from the very first real user, because the intelligence isn't something you built up over time, it's something you're routing to. This changes the calculus on what counts as "MVP-ready": the AI feature can be genuinely good in 48 hours in a way a recommendation engine or a two-sided marketplace never could be.
Pricing: a real number, not "figure it out later"
### Pricing Decision
- Free tier: 3 proposals/month (enough to feel the value, not enough to avoid paying if you send proposals regularly)
- Paid tier: $12/month, unlimited proposals
- Reasoning: anyone sending more than 3 client proposals a month is either actively freelancing or running a small agency -- $12/month is a rounding error next to what a single won proposal is worth to them
"Figure out pricing later" is one of the most common ways a weekend project never actually launches — without a price, Section 3 (Stripe integration) has nothing to configure, and the whole "shipped, payable product" goal quietly becomes "shipped, free demo." Pick a number now, even if you change it after real users respond.
You're 4 hours into your 48-hour build and realize a genuinely great feature idea — proposal templates by industry. Your out-of-scope list already includes "templates library." What's the right move?
This is exactly the moment the out-of-scope list exists for — "won't take long" is almost always wrong mid-build, and every good idea you say yes to is time taken from the core loop (auth, payment, AI feature, launch) that actually has to work by the deadline. Writing it down for later isn't losing the idea, it's protecting the constraint that makes this course's format work at all.
What you're building today
product_definition_canvas.md: your one-sentence pitch, target user, 5 MVP features, 5 explicitly out-of-scope features, your pricing decision with reasoning, and a specific success metric for the weekend (signups, or a specific revenue number) — the document every later lesson's scope decisions get checked against.