None of it is for sale

For a while the site said we sold three things: Ship, a delivery workspace; Lighthouse, a knowledge base; Buzz, a content engine that hadn't shipped past a waitlist. Three logos, three pricing pages, three onboarding flows we had to keep honest at once. None of them sold particularly well, and the reason took longer to see than it should have.

The reason wasn't the products. Ship's loop is real — the numbers on this site are pulled from it, not written by hand. Lighthouse's retrieval is real — it's still open on GitHub, still Apache-2.0, still the same engine. The reason was that we were pitching tools to people who already have tools. Every team we talked to had a stack. What they didn't have was someone who'd read their stack and could point at the expensive part.

The tell

The tell was in our own sales calls. We'd open with the loop, the DORA numbers, the seven states — and the question that came back, every time, wasn't "how much for Ship." It was some version of "can you look at what we already built and tell us what's wrong with it." We kept answering that question for free, as a warm-up to a pitch nobody asked for, and then wondering why the pitch didn't land.

We were sitting on the actual product the whole time. It just wasn't a product in the SaaS sense — it was the willingness to look at someone else's pipeline with the same discipline we apply to our own, and say the true thing about it.

What changed, mechanically

We didn't invent a new offering. We stopped selling the two things that were costing us credibility and started selling the one thing people were already asking for.

• Ship stopped being a SKU. It's the methodology — seven states, named specialists, an audit trail — written into a book and run on our own delivery every day. It still exists as a reference implementation on GitHub. Retainer and fix-sprint clients still get a workspace. Nobody signs up for it cold anymore.
• Lighthouse stopped being hosted. The paid tier is gone, the SaaS deployment is gone, the database behind it is gone. What's left is the code — open, self-hostable, the same retrieval engine, now described on the site as "the memory layer" instead of a pricing page.
• Buzz was never a product to begin with, just a waitlist page for one. It's now what it always actually was: the internal machinery that turns a finding into a field note. This post is an example of its output.
• The audit is the only thing on the site with a price. Five days, read-only access, a findings register with a dollar figure on every line. $3,500, half back if it doesn't pay for itself.

Why "none of it is for sale" is the sharpest line we've written

Every AI vendor doing an "audit" is suspected of one thing: finding problems that happen to be solved by the thing they sell. We removed that motive structurally, not just rhetorically. There's no upsell path from our audit report to a signup page, because there's no signup page. The reference implementations are free and open regardless of whether you hire us. When the incentive to pad a findings register disappears, the register gets more useful, not less.

It also means the three years of scars — the fifteen identical fix commits, the tenant-isolation bug, the lock that leaked for a day — stop being liabilities we downplay and start being the actual credential. We ran this on ourselves first. The receipts are public. That's the whole pitch now.