The quiet period has shortened
In the past, founders had months between starting a POC and filing a patent. Build quietly. Test quietly. File when ready.
Andrew's observation: that quiet period is shorter now for several reasons. AI coding tools mean POCs ship faster. Public repos and beta programs mean disclosure happens earlier. And AI patent drafting tools mean the barrier to filing a provisional on someone else's public work is lower than it used to be.
| Factor | Before | Today |
|---|---|---|
| POC timeline | Often months | Can be weeks with AI tools |
| Disclosure control | Managed release | Users, beta testers, and public repos leak constantly |
| Competitor awareness | Slow | AI tools can surface public work quickly |
| Patent drafting cost | $10k–$15k+ with a firm | AI tools can draft from existing artifacts |
None of this means every founder will be copied. It means the risk window has changed, and the old sequential model no longer fits.
Three competing demands, fixed capital
With a typical seed round ($100k–$200k), you face three competing needs:
- Build POC — engineering cost
- Provisional patent — legal cost
- Market traction — sales and marketing cost
In the old model, you would skip #2 until #1 and #3 showed traction. Andrew's view: that model assumed no one else would file on your public footprint. That assumption is riskier now.
Parallel execution: POC and provisional together
The solution is not more money. It is better allocation. Run engineering and patent work at the same time.
| Week | Engineering (POC) | Provisional |
|---|---|---|
| 1 | Build core loop, no UI | Write 1-page invention description with engineer |
| 2 | Working prototype (rough) | Draft claims around that specific prototype |
| 3 | Show to 5 friendly customers | Have AI patent tool expand description |
| 4 | Iterate on feedback | File provisional before any public demo |
| 5–8 | Keep building, now protected | Nothing — provisional is filed |
Estimated costs:
- Total legal cost: $3k–$7k (DIY with AI plus limited attorney review)
- Total engineering cost: $30k–$60k
- Time to both POC and provisional: 30–45 days
Sequential loses. Parallel wins.
Assume $100k total.
Sequential (old approach):
| Step | Cost | After 4 months |
|---|---|---|
| Build POC | $40k | Working prototype, no IP |
| File provisional | $10k | Filed, but POC has not been tested in market |
| Market testing | $30k | Feedback comes after filing |
| Remaining | $20k | Claims were guesses — no customer signal |
Problem: you filed before you knew what mattered. Your claims are guesses.
Parallel (tandem filing):
| Step | Cost | After 4 months |
|---|---|---|
| Build POC | $35k | Working prototype |
| Draft provisional alongside | $3k | Filed week 6 (parallel) |
| Market testing under provisional | $30k | Feedback before PCT decision |
| Refine claims from feedback | $2k | Claims match real value |
| Remaining | $30k | Both POC and smart filing done |
You end with more cash, better claims, and no guessing period.
Your engineer does both jobs
The biggest cost of a provisional is not the legal fee. It is the cognitive load of describing your invention clearly. Your engineer is already doing that work during POC.
| Resource | POC needs | Provisional needs | Shared? |
|---|---|---|---|
| Engineer's judgment (what is novel) | Design decisions | Invention disclosure | Yes |
| Documentation of how it works | Code, specs | Written description | Yes |
| Drawings and diagrams | Architecture | Patent drawings | Yes |
| External legal fees | No | Yes (but minimal with AI tools) | No |
You are not adding a separate expense. You are capturing the same work in two forms: working code and a written description.
Keep a running lab notebook
This approach has one requirement: your POC must be documentable. If your POC exists only as running code with no specs, diagrams, or written architecture, you will struggle to draft a provisional without stopping engineering.
The fix is simple: keep a running lab notebook. A markdown file of what you tried, what worked, and why it might be novel. That file becomes your provisional draft.
No public disclosure before filing
No Product Hunt launch. No Show HN. No conference demos. No press. No beta with strangers unless under NDA.
If you violate this, the entire strategy collapses. That is where most founders still lose — not in budget allocation, but in discipline.
Next in the series