The dream of “deployable” personal development
My former understanding of it as a lightweight playground, was clearly insufficiently, capturing its promise.
It has come along way since then I can and encapsulates an end to end flow, and going from idea-to-deployment while dictating to it, something that hasn’t existed up until now.
I am at this moment, reminded of the ( tragically!) unfulfilled potential of Google App Engine, which 15 years ago, allowed for a basic implementation of services to be written, and scaled out, with all the infrastructure building blocks present — deploying straight from your desktop/laptop to the cloud.
I feel that Replit today takes that concept, fleshes it out, and makes it even more accessible to everyone.
Removing friction
I have previously mentioned how easy it was to create live, working, examples of simple ideas I had in my head (e.g. a simple LLM wrapper app)
These are not difficult of complicated projects that we are talking about. We are not about making something that would otherwise require a large team over the course of many months (though the projects that you can accomplish in Replit certainly can fall in that range).
Rather, it is about compressing the time that you would take to build it yourself. It’s about removing friction.
Of all the mini projects that I have built in the last week or two on Replit, none are what I would describe as “very hard to do”. At the same time, I never tried any of them before, because I knew it would take at least a few days to work out all the bits and pieces of syntax one or two unfamiliar languages or environments, or the details of deployment, and getting all the pieces working, etc.
I don’t have to worry about any of that with Replit Agent. It is a delightful experience and I encourage you to experiment and see for yourself.
Not all magic
However, it’s not all black-box, and cannot be either. I recently encountered limitations — limitations with any system like this would have, where it is possible to end up in a spiraling, move off bugs, errors, and “tech debt”.
On occasions like that, you need exercise your own judgment, to figure out how fixable the situation is, and how best to guide the agent, towards what you want.
This may work — or it may not, the agent floundering while trying to respond to your demands.
As it happens, I spent a fair bit of time trying to have it first resolve the issue, then try to use or do things differently, then to go back to an earlier state and redo things from there — before giving up and simply creating a new project from scratch, with the initial prompt containing some of the dos, don’t and lessons learnt in the previous attempt.
I’m happy to say that it turned out perfectly this time round (naturally with some further “sculpting”).
The dream of “small” personal development
As I went along though, I realized something else. It’s not just about deploying things at scale. It’s equally (or sometimes more) about building for yourself.
The apps I created were either intended for me, or my immediate family to play with and use.
As an example, I’d mentioned here about a little “Quilt game” that we play while waiting for food at restaurants, with paper and crayons.
I made a digital version of it with Replit, which allows me to create and share these among us and lets us take turns uploading images.
It’s fun. No one else needs to use it. It exists, it has a URL, we can share it in Messages with each other, it’s real, for us.
That’s a lot of words for something that really needs to be experienced to be understood. So go check it out.
Originally published at http://abacusnoir.com on September 24, 2024.
