Christian Biever

Every second client brief in 2026 opens with the same sentence: "we want a chatbot." A few questions later it turns out the chatbot is not the goal at all — the goal is that people can find the right answer without knowing the exact words for it.

That is a retrieval problem, not a conversation problem. If the underlying index understands that "cancel my plan" and "how do I stop being billed" are the same intent, almost any interface on top of it feels smart. If it does not, no amount of chat polish will save you.

So I spend the first week of these projects on the boring half: cleaning the source content, deciding what a "document" even is, and measuring whether the top result is actually the right one. The chat window is the last thing I build, not the first.

In-flight Wi-Fi is a promise, not a guarantee. Somewhere over the Atlantic it will drop, usually in the middle of a push. So I stopped treating the connection as part of my workflow and started treating it as an occasional bonus.

Everything I need lives locally before I board: the repo, the dependencies, the docs, a full local build. I review, test and commit offline, and let the laptop sync whenever a signal happens to appear. The release went out somewhere past Greenland; I just didn't notice exactly when.

The lesson generalises beyond planes. An offline-first setup is simply a resilient setup — it also survives bad hotel networks, train tunnels and conference Wi-Fi, which are arguably worse than 11A.

A lounge that looks great in photos can still be useless for getting work done. After a few hundred of them, my checklist has shrunk to four things that actually matter.

Power at every seat, not one shared strip by the window. A door — or at least geometry — that keeps the gate announcements out. Wi-Fi that survives a video call. And service hours that match how people actually travel, which means coffee well after 9pm.

Everything else — the champagne, the à-la-carte menu, the view — is pleasant but irrelevant. I'd trade all of it for a quiet corner, a reliable socket and ninety undisturbed minutes.

I've been through four exits now, and none of them looked alike. One was a clean strategic acquisition; one was effectively a fire sale; one took eighteen months of due diligence; one closed over a weekend. Different numbers, different moods, same underlying lesson.

A company that depends on a single person in its founder's head is worth far less than one that runs on documented processes. The buyer is not paying for your heroics — they are paying for the parts that keep working after you leave.

So I now build as if I'll hand the keys over on day one: write things down, define who the customer actually is, and remove myself from the critical path as fast as possible. The irony is that this also makes the company far better to run while you still own it.

Automation demos love the hard case: the clever edge, the impressive end-to-end magic. In practice that 20% is where projects stall, because the exceptions are endless and the trust isn't there yet.

I go the other way. I find the dull, repetitive 80% — the copy-paste between two systems, the weekly report nobody enjoys, the renaming and filing — and I automate that first. It's unglamorous, it's low-risk, and it pays back immediately.

Once the boring work is gone, two things happen: people get hours back, and they start trusting the system enough to let it near the harder cases. The exciting 20% becomes a follow-up, not the whole bet.

Twenty years of working from the road have stripped my setup down to what fits in a single carry-on. Not as a minimalist flex — but because every extra item is one more thing to lose, charge or carry through security.

The gear is almost boring: one laptop, one good pair of headphones, the right adapters, and backups that live in the cloud rather than in the bag. The real infrastructure is the routine — fixed hours of deep work, clear handoffs, and a team that doesn't need me online to keep moving.

The hardest part was never technical. It was the boundaries: deciding when the company gets my attention and when the city I'm in does. Get that wrong and you're neither travelling nor working — just anxious in a nicer location.