Here's last week's #digital_news:
1. Let's start with Telegram, which is getting ready to launch an advanced text editor. In the iOS beta of Telegram, the developers replaced the AI editor button with a button that opens a window with Markdown markup. As promised, they plan to add a choice of text and heading styles, tables, math formulas, code blocks, block and collapsible quotes, collages, slideshows and more. Sounds like Word lite, we'll see
2. The next update is about Telegram too. Durov is preparing a major overhaul of calls. A new conference call architecture was found in the app's code. Key features:
• Switching from a regular call to a conference without hanging up
• A blockchain-based conference sync mechanism (users won't see it, but it will strengthen the infrastructure).
For now these are changes at the protocol and API level, not a finished interface. But the direction is clear: Telegram is eyeing Zoom and Google Meet territory.
3. ChatGPT will learn to talk without pauses. OpenAI is rolling out a new voice model that can listen to the user and respond at the same time. With no delays, a conversation with AI will come as close as possible to natural human speech.
4. Google has integrated computer control into Gemini 1.5 Flash. The AI agent can now control a browser, smartphone and desktop on its own: look at the screen, click, type and scroll. The company is betting on:
• Cost: Choosing the Flash model enables cheap mass automation ($1.5 per 1M input tokens).
• Safety: To prevent control from being hijacked through on-screen text, safeguards are built in: confirmation requests for important actions and an automatic stop when an attack is detected.
5. Google is also changing the rules of the game for app developers. The company is moving from a fixed fee to a flexible commission model. Developers will now be able to offer their own payment methods or send users to external websites.
6. Yandex and Neptun launched a leak protection system controlled via Alice. The key feature is deep integration with Yandex's Alice smart home. You can now shut off the water by voice or set up scenarios. For example, on the command "Alice, we're leaving", the system closes the valves by itself. What the system can do:
• Smart scenarios: You can say "Alice, I'm mopping the floor" so the sensors don't raise an alarm when water gets on the floor.
• Autonomy: It works without internet and for up to 30 hours without power (with a UPS). If the connection is lost, you'll get a notification.
• Flexibility: Sensors can be placed under sinks, washing machines or dishwashers.
7. Figma unveiled a powerful update at Config 2026. The most unusual addition is Code Layers: you can now import repositories onto the canvas and turn design layers into code and back. They also added a built-in animation tool with a timeline and shader support via WebGPU for complex visual effects. And their agents got smarter: they now work with custom scenarios and have access to external data, from Notion docs to web search
8. Anthropic launches Claude Tag, an autonomous AI agent for teams in Slack (a corporate messenger). Claude is no longer a separate window but a chat participant. You can tag @Claude, delegate a task and get on with your own work. A single agent handles tasks in a channel, so the whole team shares one history and context. The agent plans the steps of a task itself and reports back in a thread. Claude also surfaces useful info on its own and brings the team back to important issues. Data is isolated between departments, too, so sales won't see the engineers' code
Interestingly, inside Anthropic this tool already writes 65% of all code.
What do you think: is bringing AI agents into companies a passing hype or an inevitable transformation of the job market?
