Journal · Collection

Teardowns

Public AI products diagnosed with Stuck-to-Scale — the stage they are actually in, the foundation they are missing, and what we would cut first. Nobody pays us for these.

Why we publish them

The read, run in the open.

A teardown is the same diagnosis we charge for, applied to a product we have no relationship with. It is the cheapest way to show what the judgement actually looks like before anyone buys it.

We pick products that are genuinely interesting rather than easy targets, and we say what is working before we say what is not.

The format

Four questions, every time.

01Which stage is it in?The Stuck-to-Scale read: where the product sits against where its marketing implies it sits.
02What is the missing foundation?Strategy, creativity or technology — named, with the evidence that says so.
03What would we cut?Scope reduction is the default move. Usually the most contested section.
04What would we do next?One recommendation, not a menu of options.
The rules

How we keep them honest.

01Public evidence onlyMarketing, pricing, onboarding, docs and the product itself. No leaks, no insider accounts.
02No pitch attachedWe do not contact the company afterwards to sell an engagement.
03Right of replyIf a team responds, we publish it alongside — including where it proves us wrong.
04Dated and left upProducts change. The teardown keeps its date rather than quietly updating.
Recent teardowns

Three reads.

Fieldnote AI is stuck at Define, marketed like it is at Scale
The pricing page reads like a Series B company. The product reads like a validated prototype. Both cannot be true.
Ledgerly solved distribution and broke retention on the way there
A growth loop that wins the sign-up and loses the account by month two — and where the fix actually sits.
Cadence's agentic pivot skipped the thesis
An agent bolted onto an unchanged product, with no new buyer named. The rebrand is doing the work the strategy should have.
What you can do with it

Run it on yourself first.

The point of publishing the method is that a team can apply it without hiring anyone. Take the four questions to your own product and answer them in writing — most teams find the disagreement inside the room before they find it in the market.

If the answers do not converge, that is the signal the Venture Audit is built for.

You can, and we will consider it — but a requested teardown is published on the same terms as any other, including the parts you would rather we left out.

No. A teardown reads public evidence, and pre-clearing it would turn it into a marketing collaboration.

Send the argument. Right of reply is published alongside the original, and we have revised more than one conclusion that way.

In public

Our process, given away.

Written in public, revised in public — including when wrong.

Read the Journal