I don’t usually do this, but I am going to defer a controversial post for a week. I have my reasons. Singular reason, actually, and if you need to know, ask Nunya Bidness for an answer.
Vaguebooking aside, I got into a conversation on Slack today that wound up generating the response “Several People Are Typing…”
The conversation was around vibe coding and how to do low-code to no-code backend services, especially for non-engineers.
I am going to give this a name: Derploying.
Right now, I have had the pleasure of working with at least three non-technical people on AI supplemented projects.
“You are destroying the jerbs!”, someone might proclaim, but I disagree. Engineering has borne the cross of “it takes too long to get to Product Market Fit” on behalf of product management due to “the passage of time,” “we don’t have the right good ideas,” and “engineering takes time to build all ideas, even bad ones.”
The ability to use tools instead of teams to find product market fit has exposed this for what it is. A lunatic fallacy designed to shift blame.
So much for this being non-controversial.
It is really important to shorten the window to PMF, and coming up with ways to create derployments is really important.
So what can you do?
Replit mostly solves this problem, provided you do not mind being married to the Replit. What is made in Replit stays in Replit. It is hard to take stuff out of Replit. It does some crazy stuff under the hood.
If you are on AWS, you can set up a specific instance where there is a shared SSH login secret in your GitHub organization, and create a script to generate a series of project directories for you at the start of a project. Put them inside a folder for HTML deployments, and another one for server binaries, and then you will need to reserve some port numbers. That means this can only work 65,000 times before you have any serious problems. If your company is generating that many prototypes, I want to learn more. Please slide into my DMs.
If neither of these things work then it is just as easy to go grab an instance from a company like Hostinger or Digital Ocean, and just use that in the same way as the EC2 instance.
The idea is that it takes about an hour of engineering time to set up the repo, put the right derployment script in a deployment branch, and then you can move on.
If you are spending more than 300 bucks a year on this across all of your prototypes, you are doing something wrong.
If you are inside a company and freaked out by the idea of product managers pushing code to a server, don’t be. This is part of the new normal.
There are going to be two classes of product builders in the future, and by future I mean last year.
There are going to be “prototypers” and “scalers”. There will be blended engineering, product, and design percentages that will be required for each of those roles to be successful, and they will vary from person to person.
The most successful organizations out there will split their teams this way, much like teams were split up from “hunter-gatherers” to “skinner-cleaners” in their focus decades ago.
