DEV Community: Pavel Kalashnikov

Duolingo widget in 5 minutes

Pavel Kalashnikov — Mon, 12 Oct 2026 08:34:02 +0000

Built a Duolingo widget for Omalauncher (Omarchy Launcher) with Claude — right from my phone 😊

It reminds me to do my lessons. Still needs some polish, but I’ll publish it once I’ve tested it.

Or build your own in 5 minutes 😊

Omarchy on Android

Pavel Kalashnikov — Sun, 11 Oct 2026 18:37:13 +0000

This is happening on my Samsung right now 😊

Omarchy Launcher — this is going to be fun!

I’ll build a couple of widgets just for myself 😊

Two things in Rails have been seriously getting on my nerves:

Pavel Kalashnikov — Tue, 06 Oct 2026 12:48:23 +0000

The lack of a good dashboard for Rails’ default background job system, Solid Queue.
Having to jump through hoops to integrate existing dashboards into an admin panel.

So I added one directly to Tramway. The Solid Queue dashboard is now available right inside the admin area.

Tramway 4.1

☑️ Solid Queue web UI — just a few lines of code, and you can manage your app’s background jobs directly from the admin panel.

You can vibe-code your own admin panel — or ...

Pavel Kalashnikov — Sun, 04 Oct 2026 06:15:36 +0000

you can use one that’s already built, comes with its own skill for Claude and Codex, and pretty much takes care of itself.

Just shipped another small feature.

Tramway 4.0.2

☑️ Added convenient filters for index pages. Enable them with a single line of code.

☑️ tramway-skill has already been updated and knows how to use them.

https://github.com/Purple-Magic/tramway/releases/tag/4.0.2

Saving tokens also means using existing tools.

Pavel Kalashnikov — Sat, 26 Sep 2026 15:11:48 +0000

In Rails apps with Tramway and tramway-skill, this is all Claude needs to add user creation to the admin panel.

It even took the test from the example in the skill.

Ruby is more effective in terms of token safety

Pavel Kalashnikov — Fri, 25 Sep 2026 15:51:07 +0000

Purely empirically, Ruby feels way more token-efficient for programming.

I spent the last month on a JS project and the past week mostly on Ruby projects.

The difference is obvious. No further details 😄

11 years ago, we ran out of IPv4 addresses.

Pavel Kalashnikov — Thu, 24 Sep 2026 12:32:39 +0000

I wrote a short piece for the IT History Journal about how it happened — and how the internet has kept working ever since.

https://history.purple-magic.com/en/sep/24/arin-ipv4-pool-exhausted/

I'm done with framework building

Pavel Kalashnikov — Wed, 23 Sep 2026 23:19:17 +0000

With this release, I’m done trying to make Tramway a do-everything Rails library.

From now on, it’s simply a Rails admin panel you can get up and running in no time.

We’ll vibe-code the rest anyway.

Tramway 4.0

It’s been a while since I posted gem updates — and there are quite a few:

✅ Rails 7 support dropped
✅ tramway_navbar completely revamped
✅ New tramway_grid — a handy new component
✅ Updates to tramway_table and tramway_chat

https://github.com/Purple-Magic/tramway/releases/tag/4.0

Trivia night meets vibe coding.

Pavel Kalashnikov — Wed, 23 Sep 2026 22:26:42 +0000

My wife and I hosted a quiz at our friends’ wedding, but the venue had no projector.

So we vibe-coded a website instead: everyone got the questions on their phones, while we controlled the game from ours.

Claude is asking this. What’s your answer?

Pavel Kalashnikov — Wed, 23 Sep 2026 15:17:30 +0000

The first Android smartphone was unveiled exactly 18 years ago today

Pavel Kalashnikov — Wed, 23 Sep 2026 14:17:45 +0000

I wrote a short piece about it for the IT History Journal:

https://history.purple-magic.com/en/sep/23/first-android-phone-unveiled/

Free tokens!

Pavel Kalashnikov — Wed, 16 Sep 2026 10:49:26 +0000

I’ve set up a local model to use with Codex. It runs at about the same speed as Claude with Sonnet at minimum effort.

It definitely does some dumb stuff, but so far I’ve been able to fix all of it with better instructions. We’ll see how it goes.

Why did I choose gpt-oss-20b?

I had Claude benchmark around 20 models on my machine and give me a report comparing execution time and quality.

After 7 hours of testing and reasoning, it concluded that gpt-oss-20b was the best fit for my setup.

I’m going to stick with it for now and see how it performs in actual day-to-day work.