@PackBropagated As tokens grow cheaper - user attention will become a key driver.
Makes me think of the quote someone said
in a world where everyone can bake a bread, think where do you go buy your bread from today?
My take is, it would be multi agent with shared memory / data stack. Gemini and its integrations across Google surfaces is a good example.
Likely more to go into here.
Shared memory can anyway make everything feel like a single agent to the end user.
Possibly single orchestrator agents will become more popular in the future than they are today.
@durov It has been a few days since my agent switched to using this. It's been amazing - how useful pure markdown support is!!
Thanks @durov and team.
FR: any thoughts on supporting A2Ui kind of protocols?
Free market competition.
Just as it should be.
We need a plurality of AI shops for numerous reasons.
AI is the most important technology in human history.
No one company, person, or nation should dominate the exploration of this latent space.
Competition between shops will solve many problems: cost will come down exponentially; quality will continue to rise; safety and alignment will be solved; agentic features and integrations will surge.
This is one case where the supercomputer of capitalism plus free market dynamics is America's superpower.
๐ฏ very accurate take! Some of the best folks I have worked with are engineers who wear a 30% product hat and PMs who wear a 40% eng ๐ฉ.
Agency is the best skill set to look for!
Career advice nobody told you: There's nothing more valuable than someone who can just figure it out. Do some work. Ask the key questions. Get it done. Repeat. If you do that, people will fight over you.
Our mic icon just got way better on Android and iOS! This is HUGE for non-English speakers.
+ now 70+ languages
+ mix languages freely
+ don't have to change language settings
+ still doesn't interrupt you :)
Shifted my perspective on AI apps. Instead of one single "Super App" doing everything, a multi-agent setup feels way better. Let Gemini orchestrate, while hyper-specialized agents like Fitbit, Gmail, and Drive handle their verticals securely - with secure data access between them.
Incredible design pattern. ๐งฉ
@GeminiApp
even if gemini 3.5 pro is only 90% as good as the latest openai and anthropic models
that would still be acceptable
because when you pay for one of google's plans, you get a much broader ecosystem of tools and integrations:
drive storage, notebookLM, img gen, and several other multimodal capabilities
@durov Switched to using `sendRichMessage` - much better than markdownV2 - thanks ๐
On Android the markdown tables horizontal scrolls seems to be conflicting with swipe left to go back or swipe right to reply, that gets a bit annoying but acceptable.
#1 FR - native TTS for messages.
Must say this is nicely out of the box win win solution.
Better than injecting ads into the final responses. Swapping cash incentives with more tokens might be a bigger win but that'd mean ai labs taking over the ads network.
AI agents make you look at "spinners", why not get paid for it?
70% of every ad dollar, straight to the dev whose terminal showed it. We never touch your editor. Verified impressions (no abuse).
$5 and you're cashing out.
๐ https://t.co/tag4ELTT0r
We now support rich formatting for all chatbots.
Tables, nested lists, inline media, formulas, headers and more โ right in Telegram messages.
๐จ Start building! Docs: https://t.co/zgzPOOUJF5
We made scanning multiple documents super easy with the new Document Scanner in @googledrive .
Flip through the pages of a book or lay your receipts out on a table and our Document Scanner will identify, separate, and capture each page within the camera view. Scanner detects duplicates to prevent accidental re-scans automatically.
Implemented with On-Device Processing - which means, its lightning-fast and works offline.
Available to all Google Workspace customers, Workspace Individual subscribers, and users with personal Google accounts