In tech right now, everybody’s in a hurry.
That’s not because people are gullible. It’s because the pressure is real. Customers are asking about AI. Boards are asking about AI. CEOs and CTOs are making AI directives. Product teams are trying to figure out what to ship. Engineering teams can suddenly try things much faster than they could a year ago, which means they’re now being asked to try a lot more things.
The hurry makes sense, but hurry has side effects. It makes people accept more complexity and lower quality than they normally would. It makes them postpone cleanup, stack new mistakes on old ones, and enact local decisions that create global mess. When enough companies do that at once, you get today's software industry: motion, change, and no time to rethink the plumbing everything else depends on.
At Tailscale, our job is to create calm in the middle of that chaos. Our job is to be the thing people can hold onto while everything else is unstable. That means being predictable, secure, and reliable in all the right places. Put a little more dramatically: when the industry is in a tornado, we don’t help by becoming more tornado. We help by being the solid thing in the middle.
AI is useful. Judgment is mandatory.
Used well, AI saves a lot of time.
It’s good for research, debugging, drafting, explaining, and getting to a first pass faster. It can shorten the distance between understanding a problem and trying a solution, which is a big deal. A lot of work has too much blank-page time included, and blank pages are scary, and AI is genuinely good at that part.
But speed changes behaviour. When people are producing fast, and under pressure to ship, they accept more rough edges, postpone more cleanup, and make more short-sighted decisions that somebody else will have to live with later. None of that is irrational. It’s just what hurry does.
My toddler recently crossed the threshold where his head was finally big enough to fit the VR headset we have at home. He put it on and was immediately delighted by this whole new world he could look around in. Then, within about 30 seconds, he graduated from looking to flailing, walking into walls, and grabbing enthusiastically at digital sprites.
Very quickly he ran into the main limitation of virtual reality, which is reality. The headset worked. The walls were where they’d always been. But once the world got exciting and disorienting enough, somebody had to be there to make sure he didn’t walk into them.
That’s what I mean by being the adult in the room. I'm the one who gave him the headset—no, wait. Uh, I mean I'm the one who was there to stop him from bumping into more walls. That’s it.
AI is exciting enough already. Tailscale’s job is to help people use it without losing track of security, observability, compliance, or quality along the way.
The AI era needs boring foundations
From where I sit, the interesting part of AI isn’t code generation. It’s the explosion in the number of things that now can talk to each other.
More agents. More services. More internal tools. More machine-to-machine communication. More cases where something private needs to be reachable by the right thing, under the right identity, with the right policy, without triggering the Lethal Trifecta.
That’s fundamentally a connectivity and contextualization problem. Once AI leaves the demo phase and starts doing real work, it runs into ordinary reality very quickly. It needs access to private systems, but needs boundaries. It needs a trust model. It needs keys, logs, and policies. It needs … societal norms. It needs all the boring things that suddenly become very exciting when they’re missing.
The practical problem is simple: people want to connect new AI-shaped workflows to private systems without turning those systems into public ones, and without building a fresh pile of fragile networking hacks in the process.
That’s exactly the sort of boring, frustrating, high-stakes problem we've spent the last seven years working on.
Where Tailscale fits
When everything else is changing quickly, people want the underlying plumbing to disappear into the background. They don’t want to rethink how private systems connect, who gets access to what, how identity and policy get enforced, and how to do all that safely while they’re busy figuring out new tools, new workflows, and new products. The sands can't constantly shift; people are busy building on top.
This is where Tailscale shines: it’s already solved, already working right now, and not demanding your attention. Tailscale is even the thing OpenClaw recommends to make OpenClaw a little less risky (but uh, be careful with that).
As agents, APIs, and machine-to-machine workflows multiply, teams need reliable, mature containment, accountability, and auditability. The durable value here isn’t in shouting “AI” the loudest. It’s in making the underlying infrastructure simpler, safer, and easier to trust, even when you put AI on it, as thousands of companies already do.
Calm is a strategy
A lot of companies are going to spend the next few years racing to say they have an AI strategy.
That’s fine. Everyone should have an AI strategy.
But there are many kinds of strategies. Maybe your strategy is "wait and see." Maybe it's "copy the ideas that are working for people, and ignore the ones that aren't." Maybe it's "add AI to our product name for the Board of Directors but otherwise leave the product alone and see if anyone notices."
Tailscale's AI strategy is just the strategy we've always had: be kind of a late adopter. Add more nines of reliability. Offer backward compatibility forever so when something works, it keeps working. Build the safe, secure thing that helps you do what you want to do. And right now, if what you want to do is AI, well, that's great, our new tools like Aperture and Border0 mean we're here to help.
In a room full of AI experimentation, excitement, and hype, Tailscale will be there to help you, and maybe now also your swarm of thousands of agents, avoid the walls.
Avery Pennarun
