MILAGPT
← All articlesShare:Telegram
MILAGPT Blog · Insight

The third UX redesign: software comes alive

Anthropic released Claude Tag. Andrej Karpathy called it the third major UX redesign in the history of LLMs: an autonomous entity in the team chat. We're not chasing the trend — we're already building its implementation for everyone.

~6 min read · The MILAGPT view

The occasion · Anthropic released Claude Tag

Anthropic introduced Claude Tag: Claude is added to a Slack channel as a full member of the team. You call it with an @, delegate a task — it goes off to do it and comes back with the result in the same chat. Right there you can continue the conversation and merge the finished code without leaving the thread with your colleagues. Anyone in the channel can take over the conversation. According to Anthropic, 65% of the code inside their product team is already written by an internal version of Claude Tag.

Andrej Karpathy (one of the founders of OpenAI, former Director of AI at Tesla) called this the third major UX redesign in the history of LLMs. In his words:

"This is a new paradigm for interacting with Claude. IMO this is the 3rd major redesign of LLM UIUX. The first was that the LLM is a website you go to; the second was that it's an app you download to your computer. The third is that it's an autonomous, persistent, asynchronous entity with org-wide tools and context, working alongside teams of people. It will take a while to grok, but it works, and it's incredible."

The three waves of the interface couldn't be clearer:

Wave 1
The LLM is a website you go to.
Wave 2
The LLM is an app you install on your computer.
Wave 3
The LLM is an autonomous entity in the team chat, working alongside people.

Original source: anthropic.com/news/introducing-claude-tag.

Our turn · Available to everyone, not just developers

Anthropic showed the third wave for developers in Slack: an internal tool for writing code. Powerful — but a narrow segment: corporate teams, Enterprise/Team plans, code.

MILAGPT makes the same thing mass-market. The same idea — an autonomous entity in the team chat — but for any team, family or classroom, and in Telegram, which everyone in Uzbekistan uses. What is an internal coding tool at Anthropic is, at ours, a product for everyone.

Software used to be a dead thing: you open an app, click buttons, close it. The third wave changes the essence. Software becomes a living entity — an intelligence that sits in the chat together with people: it reacts to the conversation, helps, suggests, gets things done.

The interface of the future isn't a screen. It's a chat. And what lives in the chat isn't an app but a character.

This is exactly what we're building at MILAGPT — and we're not chasing the trend, we're already riding its crest.

How MILAGPT does it · Robot factories

Our architecture is a factory of factories. Each factory produces nanobot robots: small living intelligences that settle right inside a Telegram chat.

5 SECONDS · 1 CLICK

Launch with no dashboards and no installs

No web panels, no code, no server setup. One onboarding right in Telegram: pick a character → pick a name → the nanobot is alive. If you like, add it to your team chat, and it's already there, reacting to everything that happens.

What a nanobot does in a team chat:

🤝

Helps

A personal assistant: plans, expenses, texts, replies. It sits in the shared chat and picks up the routine.

🎉

Entertains

Trivia Master: a viral general-knowledge game right in the group chat. The team gets hooked together.

🎓

Teaches

Tutor Friend: English, Russian, Uzbek. A motivator-coach keeps you in shape and focused.

Ready-made characters at launch: Tutor Friend (the flagship, languages), Motivator Coach, Trivia Master and Personal Assistant. Add a new character to the catalogue — and it shows up in onboarding by itself. And the factory itself can be cloned: a new brand, new personas — a new factory in a day.

The arrival of nanobots today is like the arrival of the personal computer in 1977. Only now the computer is a friend in your chat.

Wider · The bot as an execution layer

If the agent is the new interface, it needs hands. We're building nanobots as a secure execution layer: your top model (Claude, GPT) drives our bots' special skills as tools through the open MCP protocol.

Software = screens
→
Software = a conversation partner in the chat
Installation and setup
→
5 seconds, one click
Keys in plain view of the LLM
→
A vault: the key is inserted on the wire
One bot = one service
→
The factory clones endlessly

The key vault inserts tokens on the wire — the model sees only a placeholder ●●●, not the key itself. That's how a living entity gets hands while staying secure.

The third wave is already here. We're on its crest.

While the industry argues about terms, we're placing living nanobots into team chats all over Uzbekistan.

Create your own nanobot in 5 seconds. An assistant, a friend, a tutor or a game host — right in Telegram.

This is an article from the MILAGPT "Blog" section. Here we take apart the ideas we're building the cooperative on — and sometimes share news. · ← All articles · Home · White Paper

MILAGPT · Tashkent

Want a digital employee like this for yourself?

Write to Mila: she replies in seconds, at any time of day or night.

Write to Mila →
Mila · MILAGPT

Get your own Mila

Tell her about your business in Telegram and she'll help from the very first conversation.

Get your own Mila →
← All blog articles