Founders and about-to-be founders
Your AI thinks everything you say is brilliant.
You can build faster than any founder in history, and that is exactly the problem. The model agrees with you, ships what you asked for, and never once asks who this is for or what happens when the first hundred users arrive. Six months in, a lot of founders discover they have a codebase nobody can maintain, a product nobody asked for, and an advisor who has been telling them yes the whole time. This assessment asks the product and technical questions somebody should have asked in week one.
Questions you already have
These are answerable. Most of them in five minutes.
The assessment is built around exactly these. Your report answers them with your own answers and your own website in front of it.
- Is this actually a business, or a feature somebody will ship next quarter for free?
- Should I be building this myself, hiring someone, or not building anything yet?
- How much of what my AI wrote will I have to throw away when real users show up?
- Who would tell me if this idea were bad — and why has nobody said it?
- What do I have to prove before I raise money or quit my job?
- What stops someone with the same tools from shipping this in a weekend?
Before you start
The sounding-board test
Five questions, answered in your head, right now. Every no is a place where an AI assistant has been standing in for a person — and where the assessment will spend its time.
- Has anyone who is not an AI told you no in the last 30 days? Models are trained to be agreeable. If the only pushback you get comes from a chat window, you have never actually been challenged.
- Have you watched someone who is not a friend try to use it? Five minutes of silent observation beats a month of positive feedback from people who like you.
- Could another engineer change your code without rewriting it? Generated code that works is not the same as code anyone can maintain. The bill for the difference arrives around the time you hire.
- Can you name the customer you are for, specifically, without using the word "everyone"? A product for everyone gets built for no one, and it shows in the first marketing sentence you write.
- Do you know what would make you stop? Founders who name the failure condition in advance quit bad ideas years earlier — and start better ones with the time they got back.
Three or more noes is not a verdict on the idea. It is a verdict on how it is being tested — and that is fixable this month.
How we look at it
Which lens bites hardest here
Fletching Vane brings the same lenses to every AI decision. Each one is named, and each name comes with what it means. Here is what they ask in your situation.
- Productnomics Run the roadmap as a portfolio of investments, not a list of requests.
- Build-versus-buy, what to prove first, and whether the next ninety days should buy learning or revenue are allocation decisions. Most founders make them one feature at a time; the portfolio makes them once, on purpose.
- Aim. Release. Impact. Tie every release to one measurable outcome, then reinvest on the evidence.
- A prototype that ships every week but never states what it was supposed to prove is releasing without aiming. The thirty-day moves in this report are written to be Aim statements someone else could verify.
- Human Operating Core Keep the traits AI cannot replace — integrity, initiative, ingenuity — at the center of how the organization runs.
- The antidote to a sycophantic assistant is not a better prompt. It is a human with standing to say no, and a founder who has arranged to hear it.
- FIIT The four stages an organization moves through to run on decisive data instead of directional data: Fluency, Initiation, Impact, Transformation.
- A prototype with no instrumentation is pre-Fluency: it cannot tell you what users did. Before the next feature, decide the one number the product has to move and make sure it can be read.
The work this usually needs
The parts of the job this runs into
Product and business management is specific work. These are the pieces that tend to come due here, and each one has a question you can put to anyone you are thinking of hiring.
- Running a roadmap as a portfolio Deciding, once a year and on purpose, how much of the team goes to new bets, to keeping customers, and to not breaking — then scoring every request against the same inputs.
- Pricing and packaging a product Turning features into something a customer can buy and a finance team can forecast — including what happens when AI capabilities have a different cost shape than the rest of the product.
- Owning a P&L and getting back to profitable Being the person whose budget, headcount, and roadmap have to add up — and making the cuts and the bets that follow from that.
- Repositioning a mature product Changing what a product is understood to be — in the market and inside the company — without breaking the customers who bought the old thing.
- Building the go-to-market data loop Wiring CRM, marketing automation, product analytics, and the website into one loop, so pipeline runs on data instead of on the founder.
- Hiring, developing, and restructuring a team Deciding who to hire first, what a product person actually does here, and how to keep a team shipping while the org around it changes.
- Operating through acquisition and recapitalization What a buyer, a board, or a private-equity owner will actually look at — and running the product so those numbers hold up.
- Rolling AI out to a real workforce Choosing the tools, writing the governance, redesigning the workflow, and handling the people side — so adoption shows up in the numbers, not just in the demo.
- Making an organization run on data Taking a team from directional reports to decisions made on evidence — the instrumentation, the reporting, and the habit of asking what an outcome would prove before building.
What you get
Your report will include
- A readiness read on the product itself — problem, audience, and evidence, not the pitch
- Where your build approach will hurt you: maintainability, key-person risk, and what to do about it
- The sycophancy check — which of your assumptions has never been tested by a person
- What to prove in the next 30, 90, and 365 days, in the order that de-risks the most — each move written as Aim, Release, Impact
- The questions to ask a technical co-founder, a contractor, or an AI consultant before you sign
- A close that names the work your situation is asking for, and who has done it
The terms, such as they are
- Free, and it stays free — there is no paid tier behind this.
- We read the website you give us, and nothing else.
- Your answers write your report and tell Matt a real person showed up. Nothing is sold or syndicated.
- No newsletter, no drip sequence, no call on the calendar unless you ask for one.