AI News
The latest developments, breakthroughs, and analysis from the world of artificial intelligence.
Google Launches Gemini Enterprise Agent Platform, Replacing Vertex AI
Google's new Gemini Enterprise platform gives every AI agent a cryptographic ID, ships 200+ models, and takes direct aim at OpenAI and Anthropic.
Google Splits TPUs in Two: Meet TPU 8t and 8i, Built for the Agent Era
Google's 8th-gen TPUs split training and inference into specialized chips — TPU 8t hits 2.8x Ironwood's speed, TPU 8i triples SRAM for faster agents.
Google Unveils TPU 8t and 8i: Two Chips for the Agentic Era
Google's 8th-gen TPUs split training and inference into two dedicated chips — 2.8x faster training and 80% better inference over Ironwood.
OpenAI Ships GPT-5.5: The Agent That Actually Operates Your Computer
GPT-5.5 launched Thursday with 78.7% on OSWorld, a 1M-token context window, and double the API price of GPT-5.4. Here's what's new.
OpenAI Open-Sources Privacy Filter for On-Device PII Masking
OpenAI's new Privacy Filter is an open-weight model that redacts PII locally before data ever leaves your machine — Apache 2.0, runs on a laptop.
Murati's Thinking Machines Inks Multibillion Google Cloud Deal
Thinking Machines Lab signed a multi-billion-dollar agreement for Nvidia GB300 capacity on Google Cloud, expanding a quiet 2025 partnership.
ChatGPT and Codex Hit Global Outage as OpenAI Investigates
OpenAI's ChatGPT, Codex, and API platform hit a global partial outage Monday, blocking logins, conversations, and image generation worldwide.
Claude Opus 4.7's Flat Price Hides a 20–47% Cost Increase
Opus 4.7's per-token price matches 4.6 — but a new tokenizer splits the same text into up to 47% more tokens, quietly raising real costs 20–30%.
OpenAI's GPT-Rosalind Targets Drug Discovery, Beats GPT-5.4
OpenAI's first domain-specific model targets drug discovery — and on bioinformatics benchmarks, it outperforms GPT-5.4 and every rival tested.
Sora Shuts Down in Six Days: What Users Need to Do Now
OpenAI's Sora video app closes April 26. Users have less than a week to export their AI-generated videos before the infrastructure goes dark.
Alibaba Open-Sources Qwen 3.6, a 35B Coding Model That Rivals the Frontier
Qwen 3.6-35B-A3B scores 73.4 on SWE-bench Verified with only 3B active parameters — and it's free under Apache 2.0.
Anthropic Launches Claude Design, Taking On Figma and Canva
Claude Design generates polished slides, prototypes, and one-pagers from a prompt — and exports straight to Canva, PPTX, or HTML.
Anthropic Launches Claude Design, a Direct Shot at Figma and Canva
Claude Design turns prompts into prototypes, slides, and one-pagers — and reads your codebase to match your team's visual style automatically.
Cloudflare Launches Agent Memory to Fix AI's Context Rot Problem
Cloudflare's new Agent Memory service lets AI agents remember across sessions without bloating the context window — a fix for a real, unsolved pain.
Cloudflare Launches Agent Memory, a Persistent Brain for AI Agents
Cloudflare's new Agent Memory service gives AI agents long-term recall without bloating context windows — and exports every memory back to you.
DeepSeek Breaks Tradition, Raises $300M at $10B Valuation
DeepSeek is in talks for its first-ever outside funding — $300M at a $10B+ valuation — as it preps a trillion-parameter V4 model for late April.
Perplexity's Personal Computer Turns Your Mac Into an Always-On AI Agent
Perplexity rolled out Personal Computer for Mac to Max subscribers — an AI agent that orchestrates iMessage, Mail, Calendar, and local files 24/7.
xAI Quietly Drops Grok 4.3 Beta for $300/Month SuperGrok Heavy Tier
xAI shipped Grok 4.3 beta to its priciest subscribers on April 17, with native PDF, spreadsheet, and PowerPoint generation built in.
Anthropic Starts Checking IDs: Claude Now Asks for a Passport
Anthropic quietly rolled out passport and selfie verification for select Claude users via Persona — a first among major AI labs and a jolt to its privacy brand.
Anthropic Ships Claude Opus 4.7 With Sharper Vision, Better Code
Anthropic officially released Claude Opus 4.7 on April 16, adding 3x vision resolution, an 'xhigh' reasoning mode, and stronger coding at unchanged pricing.
Avid and Google Cloud Push Agentic AI Into Professional Video Editing
Avid and Google Cloud announced a multi-year deal on April 16 embedding Gemini and Vertex AI into Media Composer and a new platform, Content Core.
Boston Dynamics Puts Google's Gemini Robotics-ER Inside Spot
Boston Dynamics is integrating Google DeepMind's new Gemini Robotics-ER 1.6 into Spot, letting the robot read gauges and reason about its surroundings.
Google Puts AI Mode Inside Chrome: Side-by-Side Browsing Goes Live
Google's AI Mode now opens web pages next to the chat in Chrome, pulls multi-tab context, and embeds directly in the New Tab page — starting today in the US.
Google's Gemini 3.1 Flash TTS Lets You Direct AI Voices With Text
Google's Gemini 3.1 Flash TTS ships 200+ audio tags, 70+ languages, native multi-speaker dialogue, and SynthID watermarking — already #2 on TTS leaderboards.