Board

So this is the problem we face. We exist because shipping software is too complicated. We are obsessed with keeping it simple and making infrastructure feel like magic. This approach has brought us hundreds of thousands of users who spend a lot of time asking for more complexity!Our answer to this and many more conundrums, is Railway v2.

It’s a tall order, but the alternative of settling for the current status quo is even more daunting.

We’ll have succeeded with v2 when two things are true:A user can come to the platform and push a button to deploy ANY service(s) or application(s) of arbitrary complexity, wired seamlessly into their existing infrastructureComplexity isn’t just “magically hidden” but provided incrementally, as layers, just in time and as the user needs it through intuitive interfaces

WebAssembly is necessary not just for its speed, but for its ability to use memory efficiently. It allows us to design highly compact data structures. Web Workers will allow our geometry algorithms to run off-thread, so that they never bog down the fluidity of the user interface.

The future of WebGPU is somewhat less certain than WASM and workers, but we will certainly be evaluating it as it matures. It offers compute shaders in its core specification, which is particularly compelling because it would allow us to offload geometry generation to the GPU.

In Andy Hertzfeld's book on the Macintosh, he describes how they’d come back into the office after dinner and work late into the night. People who’ve never experienced the thrill of working on a project they’re excited about can’t distinguish this kind of working long hours from the kind that happens in sweatshops and boiler rooms, but they’re at opposite ends of the spectrum.

Notice something in common? They all feed off of a resource that is either widely accessible or easily recreated. The Gmail API makes it easy to build an email app, the AccuWeather API makes it easy to build a weather app, Stable Diffusion makes it easy to build an AI avatar app, every programming language comes with math functions built-in, you can easily get access to people’s photos from Apple or Google, etc.To be clear, sometimes scalable businesses do emerge from categories that are usually fragmented and flavor-based. My point is that it’s rare.

Consumer packaged goods (CPG) like food, drinks, and toiletries are clearly flavor-based. But they tend to work a bit better as businesses than flavor-based software because, like media, they’re consumable: when you use them, they run out, and you need to buy more. The only difference is people are more likely to keep buying the same unique type of hand soap or whatever for a long time, whereas with media once you’ve consumed a thing you’re unlikely to consume it again. And because there are hard marginal costs associated with manufacturing and distribution, the price never goes all the way to zero in the way that it often does with software or media.

The downside is that your cost structure is probably not as great as pure tech companies, but the upside is that you have less competition.

All of these types of products can involve some flavoring to help them get off the ground. The point is not that flavoring is bad. It’s good! But it has a very specific role. Flavor provides an initial product wedge. It’s helpful for standing out initially but doesn’t provide long-term differentiation or durable advantage.

Vertical focus. Right now Lex seems like a general-purpose AI writing tool. But over time it will become more obvious that there is a subset of writing that deserves a specific tool. For example, it used to be the case that software UI designers used a design tool (Photoshop) that was made for photo editing; I think we’ll look back on today’s writing apps, like Google Docs, as similarly broad.