always bet on typescript
the trend in agentic AI tooling
Around 15 years ago, and 16 years after he created it, Brendan Eich told us to “always bet on JavaScript”.
I first learned JavaScript at an evening class in Brighton circa 2009. The web development world was emerging from the Flash days, we were all reading Smashing Magazine and web standards were beginning to cement - HTML5, CSS3 and, eventually, ES6 would follow to make the web a proper platform.
At that time, if you wanted to run some server code (like password salting or database calls) you’d use PHP (and by extension to LAMP stack).
I remember working at an ad agency in 2013 and our boss saying “let’s mess around with running JavaScript on a server - this Node.js thing looks cool”. That was madness at the time yet now is ubiquitous.
Other contenders have come along since the server-side wars to knock JavaScript off its perch: mobile apps, desktop apps, WASM, ML, embedded. Each niche has been absorbed to some extent by JavaScript - and certainly not replaced it. Not because JavaScript is the best tool for each job (the common gripe for JS haters!) but because it’s the best tool for all jobs - thanks to its place in the browser engine, ever-expanding community and gentle learning curve.
Even TypeScript itself is an admittance of defeat. JavaScript didn’t have types and that was a problem. Instead of replacing it with a strongly-typed language, TypeScript was built on top.
Since around 2015, TypeScript has become the standard. Of course it’s still good old JavaScript underneath but the appeal of TypeScript has unlocked new use cases, engineers and potential.
And now the software world is betting on TypeScript to power agentic AI. Initially, a lot of frameworks and tooling used/targeted Python, primarily a spin-off from the ML community. But nowadays, TypeScript is fast becoming the defacto language in this space.
I’ll leave you with some of the major examples:
- OpenCode
- Mastra
What have I missed?
Thanks for reading!
Log in to leave a note.