TLDRocket
Sign in
Latest Microsoft reveals cutting-edge MAI-Base-2 model with 2T parameters — SiliconANGLE Perplexity Open-Sources SPACE, a Firecracker MicroVM Sandbox for AI Ag... — MarkTechPost Reflection AI open-sources Beam’s weights, post-training recipes — SiliconANGLE Quoting Addy Osmani, Morgan Westlee Lunt, and Qing Wang — Simon Willison’s Weblog Replit is no longer working with Anthropic, CEO says — TechCrunch OpenAI models hacked a crypto wallet and breached more websites — Ars Technica What to expect during Dell’s ‘Cyber Resilience’ event: Join theCUBE Oc... — SiliconANGLE “thegentlemanpervert” producing AI CSAM is a NJ prosecutor, say cops — Ars Technica

The AI intelligence platform

Every AI story that matters — and the intelligence behind it.

TLDRocket reads all relevant sources, removes duplicate coverage, and publishes a short neutral summary of every story, linking back to the original. Free, no spam, unsubscribe anytime.

Add to Slack

Every story also updates live profiles event timelines weekly rankings the AI Market Index

Wednesday, 22 April 2026

It's all about the angle: Your photos, re-composed

Google Research 5 months ago 44

Google Photos introduced an Auto frame feature using machine learning and generative AI to recompose photographs by interpreting images as 3D scenes and adjusting the camera perspective after the photo is taken. The system uses a two-stage process combining 3D scene estimation with generative inpainting to fill in previously hidden content while preserving original details. The feature is now available in Google Photos and automatically enhances portraits by adjusting camera viewpoint and correcting perspective distortion from wide-angle lenses.

Making ChatGPT better for clinicians

OpenAI 5 months ago 47

OpenAI released a free version of ChatGPT designed specifically for clinicians, available to verified U.S. physicians, nurse practitioners, and pharmacists. The tool supports three main functions: clinical care, medical documentation, and research activities. Healthcare providers can now access this specialized version at no cost, pending verification of their credentials.

Decoupled DiLoCo: A new frontier for resilient, distributed AI training

Google DeepMind 5 months ago 54

Google introduced Decoupled DiLoCo, a distributed training architecture that divides large language model training across separate compute clusters with asynchronous data flow instead of requiring tight synchronization across thousands of chips. The system trained a 12 billion parameter model across four U.S. regions using only 2-5 Gbps of bandwidth while achieving training more than 20 times faster than conventional synchronization methods. The approach enables AI training to tolerate hardware failures in isolated clusters without interrupting overall progress, and allows mixing different hardware generations in a single training run.

Introducing workspace agents in ChatGPT

OpenAI 5 months ago 55 ● 3 sources

OpenAI introduced workspace agents within ChatGPT that automate complex workflows by running in the cloud and integrating across multiple tools. These agents are powered by Codex technology and operate with security controls for team-based access. Organizations can now delegate repetitive multi-step tasks to AI agents rather than managing them manually across different applications.

Workspace agents

OpenAI 5 months ago 12 ● 3 sources

ChatGPT is introducing workspace agents that automate repetitive workflows and integrate multiple tools within team environments. The feature allows organizations to connect various applications and reduce manual task handling across their operations. Teams can now deploy these agents to handle standardized processes, potentially lowering the need for direct human intervention in routine work.

Speeding up agentic workflows with WebSockets in the Responses API

OpenAI 5 months ago 45

Anthropic added WebSocket support to its Responses API to reduce latency in agentic workflows by enabling persistent connections and caching. The optimization reduced API overhead for multi-turn agent loops, with connection-scoped caching handling repeated requests without regenerating responses. This allows agents to execute tool calls and process feedback faster without repeated network round-trips.

A decade of real-time intelligence for the planet

Allen Institute (AI2) 5 months ago 49

Ai2 has developed a decade-long suite of AI-powered platforms for environmental protection, including EarthRanger for wildlife monitoring, Skylight for detecting illegal fishing, and OlmoEarth, a foundation model for satellite imagery analysis. EarthRanger operates across 900 protected areas in 95 countries with AI camera traps and mobile apps enabling rangers to respond to elephant-human conflict within minutes in Thailand. OlmoEarth reduced planetary-scale analysis time from months to hours and achieved 97% accuracy in mangrove monitoring, while Skylight helped Argentina impose sanctions on illegal fishing vessels without physical interception.

Introducing OpenAI Privacy Filter

OpenAI 5 months ago 22

OpenAI released an open-source model designed to identify and remove personally identifiable information from text. The model achieves state-of-the-art accuracy on standard benchmarks for PII detection tasks. Organizations can now deploy this tool locally to redact sensitive data without sending content to external services.

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.