What the shelf does to the work

For most of this year, our projects lived in separate corners. Each one had its own site, its own repository, its own quiet internal sense of how well it was doing. That arrangement is comfortable, because a project alone in a corner gets to define its own words. It decides what working means. It decides what counts as a user. Nobody is standing beside it holding a different definition.

Then you put everything on one shelf, and the words have to agree.

One page, one vocabulary

Our new Projects page asks the same three things of every entry. Who is this for. What state is it actually in. Where is the door.

The third one is the strict one. A door has to open. It is easy to write a paragraph about a product; it is much harder to put a link next to it that a stranger can click while you are asleep. Everything on that page has a destination — a live product or a public repository — and the destination is the claim. The paragraph is just the invitation.

The labels do not match, and that is the point

The temptation with a portfolio page is to flatten it. Give everything the same confident verb, the same clean badge, and let the reader assume a uniform level of maturity that does not exist.

We did the opposite. Vaktpass is live with its first association onboarded — one association, which is one, not a market. CurvLogic is an operational prototype whose commercial path is paused: it runs, and we decided not to sell it yet. Hodoscope has a waitlist but has not opened for orders. Quizt is free to play. Those states sit within a few centimetres of each other, and nothing on the page tries to smooth the difference away.

That asymmetry is the most useful thing on the page. A shelf where every label reads the same tells you nothing. A shelf where the labels disagree tells you exactly where the interesting problems are.

The open-source half has a harsher shelf

Five of the thirteen are public repositories, and a repository is answerable in a way a landing page is not. You can read it. You can install it and find out within ten minutes whether the description was generous.

So the bar for releasing one is narrow: a stranger should be able to install it, understand its boundary, and get value from it without Forge Nord anywhere in the picture. Agent Authenticator hands an AI agent a current TOTP code without handing over the seed behind it. The Proxmox Homelab Toolkit makes awkward mixed-generation hardware quieter and its sensors legible. SiteCharter Connector is the small WordPress piece that gives a site a credible way back. OpenClaw Fleet by Leassh lets you ask an agent what every machine is doing. Forge Research answers what changed, rather than what a search returned.

None of those are platforms. They are small enough to read in an afternoon, which is the only reason they are worth publishing.

Visible is not the same as validated

I want to be precise about what this page is and is not. It is not proof that these bets have paid off. Several have not been tested by anything harder than our own opinion of them. One of them we deliberately stopped short of selling. A waitlist is a promise, not a business.

What visibility buys is narrower and better than validation: it makes the work answerable. A claim with a state and a door attached can be checked. It can be wrong in a specific way, on a specific date, in front of somebody. A claim that stays in a corner is safe precisely because it can never fail.

We would rather be legibly wrong than comfortably vague. So the page is not a finish line, and I would be embarrassed to present it as one. It is a surface we can be held to.

Go and hold us to it: the projects are here. Open a few doors. If one of them does not survive contact with the thing behind it, that is worth knowing, and we would like to be the ones who find out.