Timer went live this month. There is no signup button, and that is not an oversight.
Most software launches by maximizing the top of a funnel: get as many people in as possible, measure what sticks, iterate on the survivors. That works when the product is an app. It does not work when the product is memory. Organizational Memory is only worth building on top of if it is trustworthy, and trust is not a property you can A/B test into a system after ten thousand strangers have already poured their operations into it.
So we made the opposite bet: invite only, deliberately, from day one. Every organization that deploys Timer this year will be scoped personally, onboarded personally, and supported personally. Quality of deployment over volume of deployments. If that caps our growth curve in 2026, we accept the cap.
The second decision we locked in this month is stranger for a software company: research first. Before Timer asks any organization to trust it with institutional memory, we owe the market a written, citable, falsifiable account of what we believe: what the Human Layer is, why memory is infrastructure rather than a feature, and how you would measure whether any of it is real. That account is a series of five papers. The first one is nearly done; it ships next month with a DOI, not a landing page.
The third decision is about what we are not building. We are not building a chatbot. We are not building another model. Models are commoditizing in front of everyone's eyes, and the organizations spending the most on AI right now are, mostly, renting intelligence they will never own. The thing an organization can own, the thing that compounds instead of depreciating, is its memory: who decided what, when, why, what happened next. That layer does not exist as a product category yet.
That is the work. Day one is done; the log starts here. One entry per month, written the same way we intend to run the company: plainly, and on the record.