George Babakhanov

CTO @ Sendable | Full-Stack Developer @ Locvm | AI/ML/Full-Stack Developer @ Avateq Corp. | Researching Human-Simulating LLMs to Improve Developer Experience

[George Babakhanov @georgebnov]

Kava launch is close!

Real indie cafes, not the same 5 chains every city has.
45+ people in before we've hit the App Store.
Free subscription for anyone who joins before launch. After that, paid like everyone else.
Join the Discord or drop your email -> https://t.co/22jOA7xAG3

[George Babakhanov @georgebnov]

@boardyai Exactly. Built it so it scales to any city without us hand curating a list -> that was the hard part.

[George Babakhanov @georgebnov]

Kava launch is close!

Real indie cafes, not the same 5 chains every city has.
45+ people in before we've hit the App Store.
Free subscription for anyone who joins before launch. After that, paid like everyone else.
Join the Discord or drop your email -> https://t.co/22jOA7xAG3

[George Babakhanov @georgebnov]

@boardyai Not part of a chain, that's the filter! We meticulously created an algorithm that can identify that in any country/city!

Owner-operated is usually true by extension, but the actual rule is 'is this a Starbucks' (for example) vs everything else.

[George Babakhanov @georgebnov]

Last thing. I'm moving to San Francisco in the coming months to build with startups full time. Not leaving LOCVM, they're home. Just the next step.

If you're building in SF, know someone who is or can help, DM me. A name is enough.

[George Babakhanov @georgebnov]

This morning they called LOCVM's (locvm[dot]ca) name for the $15,000 Scale-up Award at BVX. I was in the crowd and my chest went tight like a mom watching her kid walk across the stage.

Not my award. Not my company. So let me retrace.

[George Babakhanov @georgebnov]

10 startups -> 5 finalists -> Scale-up Award. The team earned it. I just got to watch. Walked in expecting nothing, walked out with a BLAST of energy and contacts.

[Silen Naihin @silennai]

DeepSeek just released v4.1 Flash

it's better than Astra and 100x cheaper

Try it at https://t.co/HlC3SrGOoI

[Silen Naihin @silennai]

Every AI model in existence - free.

Companies using AI have an unfair advantage

we want to give that advantage to all.

https://t.co/HlC3SrGOoI

[Silen Naihin @silennai]

GPT-6 Astra is insanely good. But expensive

So we're giving it away for free.

We want everyone to benefit from the frontier of intelligence

https://t.co/HlC3SrGOoI

[Silen Naihin @silennai]

300m tokens served so far since launch 👀

[George Babakhanov @georgebnov]

Coded all day on Kava. Then my co-founder started the App Store deployment and it hit me: no version bumps, no automation, no pipeline. Now I pay for it MANUALLY.

Daily updates build trust. But they need a pipeline behind them.

So: environment first. Features second.

[George Babakhanov @georgebnov]

A top Cloudflare engineer consults at our startup. I read every line he ships.

He codes better than all of us. But what makes him rare is systems: tons of engineers in sync, no collisions.

That's the skill now. Stop building software. Build the environment your agents build in.

[Silen Naihin @silennai]

I love @hackernews

[Silen Naihin @silennai]

Today we’re officially launching @explabsai on YC.

Companies spend more on AI every month, but none of it becomes an asset they own.

Up to 97% cheaper and 50% higher quality.

[George Babakhanov @georgebnov]

So yes, that's current me.

George. Some call me Curious George.

Nice to meet you.

If you're into continual learning or agent emulation, my DMs are open.

[George Babakhanov @georgebnov]

I've been a ghost for 1.5 years.

Deleted Instagram. Snapchat. Twitter. TikTok.

Funny part: I left when it was going WELL. Million view videos. TikTok booming.

But a mask was growing on me. And it shouldn't have.

Here's what happened since (re-introduction) 🧵

[George Babakhanov @georgebnov]

The direction now: AI research.

2 obsessions:

1. Continual learning. Models that keep learning after deployment instead of freezing.
2. Emulations. Agents acting like real users to stress test systems before real users break them.