This media is not supported in your browser
VIEW IN TELEGRAM
Google is rolling out AI task automation in Android. On Pixel 10 and upcoming Samsung Galaxy S26 devices, Gemini will be able to open apps and perform actions on behalf of the user.
A user can ask Gemini to order an Uber to the airport. Gemini launches the app in a virtual window, selects the ride details and prepares the booking, with final confirmation left to the user. If choices are needed, it asks. At launch, Uber and Grubhub are supported.
The feature will become part of Android 17. Gemini 3 powers the system, navigating app interfaces through reasoning and simulated clicks, with optional developer frameworks to streamline integration.
Please open Telegram to view this post
VIEW IN TELEGRAM
๐คฉ61โค57๐52๐ฅ52๐31๐ค9๐3๐ข3๐3
This media is not supported in your browser
VIEW IN TELEGRAM
AiCandy, founded in 2025, broke out after two Belgian creators released a 40 second AI generated ad.
The video aged Elon Musk, Jeff Bezos and Sam Altman and promoted โEnergym,โ a fictional 2036 fitness club where unemployed people pedal to power the AI that replaced them.
The clip reached 5M views across social platforms. A U.S. senator commented on X that it โdoesnโt look like a parody.โ
Within a day, the creators received a job offer from The Dor Brothers and multiple collaboration requests.
One short AI video pushed a small studio into global feeds.
Please open Telegram to view this post
VIEW IN TELEGRAM
โค100๐72๐35๐ฅ32๐ข16๐ฆ11๐6๐3
Please open Telegram to view this post
VIEW IN TELEGRAM
๐88๐ฅ49๐ณ43โค40๐28๐4๐ค3๐1
Please open Telegram to view this post
VIEW IN TELEGRAM
๐75โค70๐55๐ฏ52๐ฅ7๐2๐ข1
Please open Telegram to view this post
VIEW IN TELEGRAM
โค138๐คฏ67๐50๐ฅ24๐ข2๐1๐1
A new benchmark measures how often language models recognize nonsense in a prompt instead of going along with it.
The chart shows three outcomes: green means the model flagged the issue, yellow means partial objection, red means it complied.
Most of the Claude family scored highest on detecting flawed or misleading prompts.
Please open Telegram to view this post
VIEW IN TELEGRAM
โค123๐ค116๐ข3๐1๐1
This media is not supported in your browser
VIEW IN TELEGRAM
In 1986, Casio released the Pocket TV-200, a portable television that ran on 2 AA batteries for about 2โ3 hours.
The device used a natural light system. Light passed through a monochrome LCD matrix, and the final image was reflected onto a mirror inside the unit.
Long before smartphones, Casio was already experimenting with handheld screens powered by simple hardware.
Please open Telegram to view this post
VIEW IN TELEGRAM
โค143๐113๐ฅ4๐3๐3๐ข1
Anthropic introduced a simple way to move long-term memory from another AI into Claude. The update comes as more users switch from ChatGPT to Claude.
Users can export saved context from their previous assistant and paste it into Claudeโs memory settings. The process takes about a minute and avoids re-explaining preferences or project details.
Anthropic provides a ready-made prompt to use in ChatGPT or Gemini, then asks users to copy the response into Claude.
Please open Telegram to view this post
VIEW IN TELEGRAM
๐80๐ฏ55โค53๐ฅ30๐7๐ข4๐3๐1
Startups & Ventures
This media is not supported in your browser
VIEW IN TELEGRAM
The mysterious story around OpenAIโs headphones didnโt end with that video. Now Airbnb co-founder and chief designer Jony Ive has โaccidentallyโ appeared in them.
๐ @tech
Please open Telegram to view this post
VIEW IN TELEGRAM
โค109๐103๐12๐ฏ6๐ข4๐ฅ1๐1
This media is not supported in your browser
VIEW IN TELEGRAM
Voice mode is now live in Claude Code for about 5% of users, with a broader rollout planned in the coming weeks. Access appears via a note on the welcome screen, and can be toggled with
/voice.It works as push-to-talk. Hold space, speak, release. The transcript streams directly at the cursor position, so users can mix typing and voice in the same prompt.
Voice mode is included at no extra cost. Transcription tokens do not count against rate limits. It is rolling out across Pro, Max, Team, and Enterprise plans.
Please open Telegram to view this post
VIEW IN TELEGRAM
๐96โค87๐ฅ37๐ฏ17๐1๐ข1๐1
Apple refreshed the 14-inch and 16-inch MacBook Pro with new M5 Pro and M5 Max chips built on a dual-die Fusion architecture. The focus is AI performance and higher memory bandwidth.
Key specs:
Prices start at $2,200. Preorders open March 4, sales begin March 11.
MacBook Air also moves to M5, with 512 GB base SSD, up to 4 TB storage, 153 GB/s memory bandwidth, and a starting price of $1,100.
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
5๐คฏ61๐ฆ44๐43๐42๐ค42โค40๐ฅ36๐ข34๐33๐33๐ฏ26
Media is too big
VIEW IN TELEGRAM
OpenClaw demonstrated real-time 3D reasoning on the Unitree G1. The system builds dynamic voxel maps to track objects and temporal changes.
It embeds spatial data for deep queries, making it suited for drones and quadruped robots.
The future of perception.
Please open Telegram to view this post
VIEW IN TELEGRAM
๐100โค86๐ฏ78๐ข4๐2๐1๐ฅ1
Media is too big
VIEW IN TELEGRAM
This is Appleโs biggest push yet into low end laptops.
Available March 11th.
Please open Telegram to view this post
VIEW IN TELEGRAM
๐ฅ79โค57๐57๐ฏ56๐4๐2๐1
Reports say OpenAIโs upcoming GPT-5.4 could bring a context window of up to 2M tokens along with persistent memory. That would allow the model to work with large codebases and long document sets without constant re-prompting.
The model is also expected to support full-resolution image processing for PNG, JPEG, and WebP files. That matters for dense screenshots, technical diagrams, UI layouts, and documents with small text where downscaling usually loses information.
Another addition is a new speed-priority tier plus an โextremeโ reasoning mode. This mode reportedly allows the model to spend more compute on difficult tasks, aimed at research workflows and long multi-step jobs.
Mentions of GPT-5.4 briefly appeared in pull requests inside the public Codex repository before being removed. Prediction markets currently assign a 55% chance of release before April 2026 and 74% before June.
Please open Telegram to view this post
VIEW IN TELEGRAM
โค102๐ฅ90๐79๐9๐4๐1๐ข1
A report by Svenska Dagbladet describes how footage from Metaโs AI smart glasses is reviewed by human data labelers. Journalists visited a Sama office in Nairobi, where workers analyze videos to help train Metaโs AI systems.
Workers say they watch recordings captured by users in everyday situations. The tasks include labeling objects in videos and checking chatbot responses. Some clips contain sensitive scenes, including private moments at home or visible financial information.
The investigation also looked at how users can control their data. The glasses require internet connectivity because AI processing happens on Metaโs servers. Journalists say they could not get a clear explanation from Meta on how users can fully prevent their videos or conversations from being used.
The report suggests that most users would likely keep using the device even if they knew their footage might be reviewed.
Please open Telegram to view this post
VIEW IN TELEGRAM
๐86๐76๐ฏ71โค38๐15๐2๐1๐คฏ1๐1
Please open Telegram to view this post
VIEW IN TELEGRAM
๐ฅ96๐68๐ฆ68โค50๐2๐ฏ2๐1๐1
This media is not supported in your browser
VIEW IN TELEGRAM
Google introduced a new feature in NotebookLM called Cinematic Video Overview. It turns user sources into animated video explainers instead of static slides.
The system combines several models. Gemini 3 acts as the director that structures the story. Nano Banana Pro generates visual references. Veo 3 produces the animated video.
The feature is currently limited to Ultra accounts.
Please open Telegram to view this post
VIEW IN TELEGRAM
โค76๐72๐ฅ58๐57๐ฏ6๐ข4๐1