Anushervon Niyozov

AI & Full-Stack Developer
Building AI products, automations & web apps
Next.js • Python • FastAPI • Supabase
SWE @ Iowa State ’29

[Anushervon Niyozov @Niyozovv]

I’m Anushervon — a Software Engineering student at Iowa State building AI products and full-stack software.

I work with Next.js, Python/FastAPI, Supabase/Postgres, and LLM integrations.

I’ll use this account to document what I build, break down technical problems, share experiments, and publish what I learn while shipping real products.

Currently building:
• https://t.co/FRSwuQyPbb — automotive marketplace
• Ansar — AI-powered intelligence/news platform
• Creative OS — AI/software project

I’ll share the wins, mistakes, architecture, code, experiments, and lessons along the way.

Building in public from here.

[Anushervon Niyozov @Niyozovv]

@iam_3bu1 That makes sense. I run into this constantly with Claude Code. Even simple commands to read files can interrupt the entire workflow until I approve them. A few hard boundaries with more freedom inside the project sounds like a much better balance

[Anushervon Niyozov @Niyozovv]

The hardest part of building an AI agent that uses your computer isn't making it click buttons.

It's knowing when it should stop and ask for permission.

Sending messages, deleting files, submitting forms — these are real actions with real consequences.

An AI agent should save you time, not make you nervous about what it's doing.

[Anushervon Niyozov @Niyozovv]

@iam_3bu1 Exactly. And I think the tricky part is avoiding approval fatigue. If an agent asks for permission every 10 seconds, it stops being useful.

The challenge is finding that balance between autonomy and trust.

Would you keep those four rules fixed, or make them customizable?

[Anushervon Niyozov @Niyozovv]

@elonmusk

[Anushervon Niyozov @Niyozovv]

I’m building tiny tools to make the internet a little less annoying.
What’s one small online task you wish took 10 seconds instead of 10 minutes?
I’ll pick one reply, build a mini-tool for it, and share the experiment including what fails

[Anushervon Niyozov @Niyozovv]

@DensedD1320 @thsottiaux So true

[Anushervon Niyozov @Niyozovv]

Creativity has always set humans apart. We went from primitive tools to transmitting information across the planet through invisible waves. Nothing else on Earth has achieved anything like it.

The problem with AI in design has been its tendency to reproduce what already exists rather than create something genuinely new.

If that limitation is being overcome, the implications are enormous. Within four years, a single person could potentially build a company with enterprise-grade software and design using just a few inputs.

That raises a bigger question: how far will this technology go, and what will ultimately limit it?

I'm going to test whether this is a real breakthrough or just another impressive demo.

[Sam Altman @sama]

ChatGPT can now generate a custom UI for you!

[Anushervon Niyozov @Niyozovv]

We didn’t place. The product worked. The presentation didn’t. Here’s what we built, what we got wrong, and what we’ll do differently next time https://t.co/RvjWKgltAG

[Anushervon Niyozov @Niyozovv]

@Ammaryasser404 Hehe nope, we COMMENT via # directly in a code to remember in future what it does. But if you lazy, AI helps in understanding, editing much more 👍

[Anushervon Niyozov @Niyozovv]

This style 😁🔥