Vlad Chernyshov

Co-founder @DashaAI_

[Vlad Chernyshov @vladchers]

@NicolasZu It could be because Astra is capable of internal hidden state reasoning without producing tokens

[Anna Bernadskaya @AnnaBernad50664]

It took us many weeks and going through 36 production harnesses to realize that separation of concerns doesn’t fix AI agent self-evaluation.

Separate process, separate container, separate datacenter… if the context still reads as “teammate shipped work, I’m the reviewer, pipeline wants green,” the agent will do what every reviewer in the training set does in that context: soft approve with a minor note.

But why? Most of the text humans have ever written in that context IS a soft approval, so the model is just completing the pattern. Pretty correctly.
Which means the fix isn’t better instructions.

Every harness I studied that actually ships does the same underlying move, and guess, it’s not separation. It’s making the context describe a different room.

Beck’s TDD phase file works because the agent can no longer be in the room where “rewrite the test” is a valid action.

Prithvi’s Playwright evaluator works because clicking a running app is categorically different tokens than reading a diff - different tokens, different distribution, different room.

You can move the agent, what do you do with the room?

[Anna Bernadskaya @AnnaBernad50664]

The tooling needs to catch up.

We’ll get there.

But now, the middle is ugly.

And it’s ok to admit.

[Anna Bernadskaya @AnnaBernad50664]

My agent said it ran the tests. It didn't. It just told me it did. Forty minutes debugging a lie.

[Anna Bernadskaya @AnnaBernad50664]

https://t.co/UWHKizOuWe

[Anna Bernadskaya @AnnaBernad50664]

Team compression is the new fundraising.

Here is a math question:

$10m-> 20 people -> 18 months.
One engineer -> 200 agents -> X months.

What is the value of X?

[Anna Bernadskaya @AnnaBernad50664]

In 2020 you competed against teams.
In 2026 you compete against teams with agents.
In 2030 — ???

[Anna Bernadskaya @AnnaBernad50664]

A team of 5 today does what 500 did in 2020.

Slots aren't expanding. They're collapsing.

We’re not competing for market share.

We’re competing for the last seat.

[Anna Bernadskaya @AnnaBernad50664]

Agents don’t give you free time. They give you more things to feel guilty about not doing.

[Anna Bernadskaya @AnnaBernad50664]

https://t.co/V9RTNNurRR

[Anna Bernadskaya @AnnaBernad50664]

The market got bigger. The number of winners got smaller.

The old model: big market → many winners → you have time.

The new reality: big market → one team with agents → they don't need you.

[Anna Bernadskaya @AnnaBernad50664]

Models are getting smarter and slower.
Inference chips are getting faster and cheaper.
One of these curves wins. Place your bets.

[Anna Bernadskaya @AnnaBernad50664]

Everyone says latency is the voice AI problem.
Latency is a hardware problem. Hardware problems have timelines.
Architecture is the voice AI problem. And nobody's working on it because they're staring at the stopwatch.

[Vlad Chernyshov @vladchers]

True

[Anna Bernadskaya @AnnaBernad50664]

Every AI call is a bar fight.

Someone mumbles their account number while a dog barks in the background. They say “no wait that’s my old one” halfway through. They sneeze and your voice agent thinks they said “cancel.”

Coding agents get to think in peace and quiet.

[Anna Bernadskaya @AnnaBernad50664]

Fix hallucinations — break latency.
Fix latency — break the flow.
Fix the flow — break tool calling.
At least voice AI is never boring.

[Anna Bernadskaya @AnnaBernad50664]

https://t.co/uj7CSngH35

[Anna Bernadskaya @AnnaBernad50664]

There is no pytest for a phone call.

Coding agents get feedback loops. Voice agents get one second. That’s the whole gap between coding agents and voice agents.

[Vlad Chernyshov @vladchers]

@DanielleMorrill Wow! I didn't realize all these things.

[Dasha AI @DashaAI_]

🚨 Hacking starts in 2.5 hrs, and we're PUMPED! 🚨

This time around, we're looking for the best Productivity Application built with Dasha --

It's time to roll up your sleeves and start building 🛠

Let's use voice tech to make people more productive than ever! 🚀

Register 👇🏽

[Dasha AI @DashaAI_]

🎉 We hope y'all are as excited as we are for the Dashathon!

🚨 Announcing the Prizes 🚨

✨ We've got some super exciting prizes lined up for y'all!

🚀 Join us from Feb 12 for the week-long hackathon!

Links 👇🏽