Twenty Years of Brunello: The Audit We Owe Ourselves

We turned twenty this month. The first post, in March 2006, argued that building websites on web standards was worth the extra trouble. This month we spent part of a Tuesday watching an AI agent complete a checkout on a client’s site.

That is the distance twenty years covers in this business. And because we have published our thinking the whole way, every call we made is sitting right there with a date on it. So instead of a celebration post, we are doing the braver version: an audit. What we got right, what we got wrong, and what actually held up.

What we got right

Four bets have paid off consistently enough to call them the through-lines of the whole twenty years.

Standards over shortcuts. Semantic, well-structured sites held up through the mobile shift, the accessibility era, and new device categories nobody planned for. Now that same structure also helps AI systems understand a site.

Owned audiences over rented reach. We started making this argument when Google Reader shut down in 2013, and every platform shake-up since has confirmed it. The email list and the website you control outlast every algorithm change and every acquisition. This is the single most valuable piece of advice we have ever given.

Treating announced deadlines as project calendars. When a platform tells you what it will enforce and when, believe it. Clients who took that approach met every mobile, security, and performance deadline of the past decade calmly. Usually months early.

Engaging new technology without buying the hype. The same sorting question every time: what job does this actually do better, for whom, and at what cost this year? It told us to build on AJAX in 2005, to wait on chatbots and web3, and to take generative AI seriously the week it arrived.

What we got wrong

An audit that only finds wins is a brochure, so here are the misses.

We tolerated separate mobile sites for too long after responsive design had clearly won. Some readers heard tolerance as permission, and it cost them later.

We repeatedly underestimated how long old versions of Internet Explorer would linger in large organizations. We were modeling software release notes when we should have been modeling procurement cycles.

And our early enthusiasm for structured data was right about the destination and a full decade wrong about the arrival time. The pattern across all three misses is the same. Our errors were timing errors, made whenever we let a correct principle stand in for an honest schedule. Principles can tell you where things are heading, but not when.

What we would tell our 2006 selves

If we could send one page back to the two people setting up that first office, it would say something like this.

The technology you are agonizing over will not matter in five years. The habit of choosing it carefully will matter forever. Write down why you made each decision, because the reasons outlast the decisions.

Answer the email from the small nonprofit. That relationship will outlive three of your biggest contracts, follow two people through four job changes, and eventually bring you a client you could never have pitched.

And keep publishing, even in the months when it feels like nobody is reading. Especially then. The posts you write when a technology is new and uncertain become the proof that you were paying attention.

What actually held up

Underneath the wins and losses sits the thing that made twenty years navigable, and it is not a technology position.

Technologies change faster than the reasons institutions need them. What clients have paid us for, through every cycle, is judgment. Which changes matter for their actual mission. Which claims survive their own demos. Which investments will still be paying off when the conference that launched them is forgotten.

The tools for that judgment are simple and repeatable. Write down requirements before watching demos. Grade vendors on how they have treated their existing customers. Ask what leaving would cost before signing anything. Check your own past calls honestly, the way this post is trying to do.

Thank you for checking our work

Publishing everything, mistakes included, built the only kind of credibility we ever wanted: the checkable kind. The readers who checked it, argued with it, and told us when we missed are part of whatever we got right.

The founding post asked whether doing things properly was worth the trouble. Twenty years later the answer turned out to be bigger than markup. Doing careful work in public, in writing, compounds. It compounds fastest when the industry is loudest, and the industry has never been louder than right now.

Year twenty-one’s first posts are already drafting, and they will be held to everything this one just promised. Thank you, sincerely, for twenty years of reading.