Automation

 

I’ve been messing around with AI coding assistants since the very beginning; back when they were nothing more than "autocomplete" tools and couldn’t actually write full code. Watching them evolve has been wild. Now, in many cases, they can write faster and more accurately than most human developers. Still, they need guidance. Using AI feels a lot like working with mid-level developers; you still have to point them in the right direction, and they don’t really have the depth of real-world experience.

With AI, I can juggle three or more projects at once. Research that used to take days? a few seconds to a few hours. Syntax errors? Almost never. But it’s exhausting. When I lead multiple projects with human teams, we take our time; research, development, mid-project checkpoints every few days. With AI, things move so fast I’m checking in multiple times a day. Sometimes it just gets stuck; it struggles with creative solutions, and I have to step in and guide it. Even so, it’s like having 2-3 developers on each project. Running three projects at that pace will wear anyone down.

I’ve seen a lot of claims that this process can be fully automated; that humans barely need to intervene. My goal is to get close to that, so I can sit at a single screen, approve things, guide the AI, and skip hopping between multiple IDEs and terminals. It reminds me of managing human teams; early on, I’d walk around, hold multiple meetings, check in constantly. Later, I could step back and rely on reports and high-level updates. AI seems to be following a similar path; starting as a junior developer, moving to mid-level, then senior, and eventually lead. Hopefully, one day it’ll reach architect-level thinking and maybe even handle some managerial tasks.

I’m excited and also worried. It will definitely eliminate many jobs. But at the same time, we can’t stop this evolution. Just like the industrial revolution in the past; people smashed the machines and all, but eventually every factory used them.

And AI amplifies our ability. Imagine this: without computers, power tools, and machines, we couldn’t build cars, airplanes, or rockets. With AI, we can reach much higher goals.

On the business side, I can see AI replacing roles like accounting, HR, marketing, and sales sooner rather than later. Who knows; maybe one day an entire company could run almost entirely on AI.

Backup using Restic

I used to rely on my own shell scripts to back up different directories and files scattered around my system. They worked, but I was always tweaking them whenever something new came up. A while back, I discovered a tool called Restic and ran it in testing for about a year. It’s worked really well, so I’m writing down my setup here. Restic does basically the same job my scripts did, but better ad more. My scripts kept a list of files and directories, and crontab would use that list to run the backups. Restic has features like remote server support, encryption, snapshots, and more.

Flow / Design

Two crontab steps with separate schedules:


### 1. run restic backup with file list

```mermaid
flowchart LR
    A[Crontab Trigger] --> B

    subgraph B[Restic Backup Script]
        C[Restic Backup Uses Files List]
        D[Restic repository on backup server\nusing a VeraCrypt-encrypted disk]
        C --> D
    end


```

### 2. run cloud backup

```mermaid
flowchart LR
    A[Crontab Trigger] --> B

    subgraph B[Cloud Backup Script]
        C[Unmount VeraCrypt Images]
        D[Rclone Copy VeraCrypt Images to Cloud]
        E[Remount VeraCrypt Images]
        C --> D --> E
    end


```

I haven't written a blog post in a while. I've been busy learning about AI, specifically LLMs, and building software/system with the help of AI coding assistants. There's no need to talk about how amazing it is, how much has changed in such a short time, or how quickly things keep evolving; it's honestly impossible to keep up.

I usually write blog posts about things that took me a long time to figure out, hoping others might benefit from the same experience. But with AI, it feels like anyone can figure things out pretty quickly, so I started to wonder whether writing a blog was even useful anymore. After spending more time with AI, though, I realized it still lacks clear direction and a solid high-level perspective. And besides, a "blog" is really just a log; it doesn't have to be perfectly polished or full of detailed, practical information. With that in mind, I've decided to keep writing occasionally about what I'm working on, simply as a record.  Maybe some findings too.