Hello, little ones!
Today I want to offer you a completely new experience in installing, configuring, and interacting with a personal AI agent. Just apply three times a day... go to Gigachat (suddenly) and send the link: https://github.com/Skunokil/vosmenog/blob/main/INSTALL_PROMPT.md and then just follow the instructions.

The longread below is for those who don't understand why they need it or if they need it at all.

As far as I understand, the world has finally and irrevocably entered an era where a personal J.A.R.V.I.S. will appear in everyone's phone, not because it's fun and trendy, but because otherwise you won't be able to study or work fully. (Traditional professions are also affected).

What a twist? We can discuss in the comments. But at least I'm tired of searching for experts on my own issues, writing them motivating letters, finding time in their work-life balanced calendars, and in the end getting a pile of counter-questions for clarification instead of expertise. That's the first thing.

Second. Neural networks are also quite the chatterboxes. But not only are they chatterboxes, they are also energetic idiots. You can listen to stories about the brilliant Claude Code as much as you want, but those who have tried to do something with it know what it's like to "eat crap".

Third. Money. I'm very happy for people who can buy subscriptions for $2K and sit generating brilliant code and neural slop. My moral limit is $20, because I'm an amateur. In the sense that I'm neither a designer, nor a blogger, nor a developer. Or anything really.

So, in short, the situation: on one hand, I need to use advanced models in large quantities at work to not lose my job; on the other hand, I don't want to give my salary to overseas capitalists from Anthropic. (All according to TRIZ canons)

I tried working in different ways.
- Brainstormed in Perplexity, then coded in Claude Code
- Moved brainstorming to Claude, and became its assistant in the Linux console.
- Experimented with technical and product documentation and invented versioning on GitHub
- Installed OpenCode and entrusted it with periodic document and code audits.

But there was still a lot of suffering, and limits were low. Problems with logging, context memory, and stable behavior remained. And to be completely honest, the code turned out mediocre, and solutions couldn't meet users.

What do you think, why haven't I written for so long? That's right!

But finally a breakthrough happened.
Vosily has acquired a framework and an open-source body for an AI agent.

You can find out all the details after entering the prompt (see spoiler), but essentially I suggest you don't waste months on desperate experiments and suffering, but instead install an open-source agent with:
✍️ - Learning mode,
👨‍💻 - Preconfigured combined plan/build behavior,
💡 - Memory,
👀 - Documentation skills,
😃 - A hard-earned work framework.

And most importantly, all of this is completely free. Because our Chinese brothers allow unlimited use of their GLM-5.2 under the hood of Opencode. And it's really a powerhouse (although without the framework it can create a real mess).

And Claude now also documents threads according to this framework. And it is used for really complex tasks. And the limits for $20 are more than enough.

So, brothers and sisters, welcome to the world of personal agents, available to every housewife without VPN or paid subscriptions.