As a professional developer with tons of projects, the problem of deploying them into the real world becomes apparent. To me it came at the time I bought Sber.tech, the project was such a pity to deploy: build docker, run commands to setup database, configure nginx.

I started seeking for an easy solution and found coolify.io. With coolify there are just two problems:

compute requirements. Coolify brings in couple of docker containers, hefty docker images, uses traefic instead of nginx as a reverse proxy. It is suboptimal.

Second is low accessibility in terms of developer experience. I want to understand what is going on within the system, want to write templates and plugins or completely rebuild platform to fit my needs.

Principles of Jakeloud, can be extracted from pains above:

That’s why in Jan 2023 I started Jakeloud. Summer-ish it became viable solution, running all my apps.

Next big thing to automate is email server and intranet, that creates company in a click. Make something competitive with google suite (Workplace) or Yandex suite (360), eliminating friction, only leaving important services and interconnecting them. This part is unclear, because company should be quite developed to actually feel any difference from these automations.

Another stretch goal is *sigh* generative AI. Okay, listen, this is not about hype. Neural networks can already paint, understand guess how to build product. Let’s just start by building pure html page from text description. This task will certainly be valuable, as there are 1-2 feasible services that successively solve it. And after that, I hope architecture will be laid and adding JS or backend functionality will just cost some data mining and machine learning.

Think of it from another perspective. At the start of programming you had to do everything by hand: write assembly on paper. Then interpreters and compilers came along, that reduce ops. Then we got IDEs, automated testing. Then CI/CD, linting, e2e tests came along. Now we got even upkeep automated (so called status pages, RUM). We can see a trend that ops gets automated first. So moving from automating ops to automating some programming looks promising. That’s why Jakeloud would feel in its own cup of tea with generative AI. Automating devops is much easier, it has more in common with present models: you have to have surface knowledge in many different areas.