web3 and ai trends / open source /
Anthropic's new Code Review feature deploys a team of parallel AI agents on every pull request to catch logic errors before they hit production, targeting enterprise teams flooded by AI-generated code.
Elon Musk announced 'Macrohard'—a joint Tesla-xAI AI agent project—just hours after Business Insider reported the effort had stalled internally amid leadership turnover.
Microsoft's 'Wave 3' Copilot update introduces Copilot Cowork, a multi-step AI agent built in collaboration with Anthropic that autonomously executes tasks across Outlook, Teams, Excel, and PowerPoint.
OpenAI is reportedly developing its own code repository platform after repeated GitHub outages left engineers unable to commit code for hours at a time.
Ethereum co-founder Vitalik Buterin warns that AI agents holding crypto wallets are vulnerable to prompt injection attacks hidden inside ENS profiles — a threat that no current solution fully solves.
Meta revealed four new MTIA processors built on RISC-V architecture and manufactured by TSMC, designed to power AI inference and recommendation systems across Facebook and Instagram.
Amazon imposed new code-review safeguards after a series of e-commerce outages, with at least one disruption traced to its own AI coding assistant Q.
Amid its legal battle with the Pentagon, Anthropic launched the Anthropic Institute — a new internal think tank of 30 researchers focused on AI's impact on jobs, economies, and democratic control.
CENTCOM commander confirmed AI tools including Palantir's Maven Smart System—built on Anthropic's Claude—are accelerating targeting processes in Operation Epic Fury against Iran.
Niantic Spatial is putting its 30-billion-image map of the world — built from years of Pokémon Go players scanning city streets — to work inside Coco Robotics' sidewalk delivery fleet.
NVIDIA and Thinking Machines announced a strategic infrastructure partnership: a 1-gigawatt DGX Cloud cluster and early deployment of NVIDIA Vera Rubin systems for a new frontier-model platform.
Cortical Labs launches the world's first Biological Data Centre in Melbourne with 120 internet-connected CL1 units — and developers are already using living neurons to play DOOM through the Cortical Cloud API.
Mind Robotics, founded by Rivian CEO RJ Scaringe, closed a $500M Series A co-led by Accel and a16z, reaching a $2B valuation as it bets on AI-trained industrial robots over humanoids.
Apple's M5 MacBook Air and M5 Pro/Max MacBook Pro went on sale today, featuring Neural Accelerators in every GPU core and a claimed 4× boost in AI compute over the previous generation.
Nvidia launched Nemotron 3 Super, a 120-billion-parameter open-weight model delivering 5x higher throughput for multi-agent AI systems, alongside a revealed $26 billion five-year investment in open-weight AI development.
Galileo's new Agent Control platform lets enterprises write guardrails once and deploy them across all AI agents, released under Apache 2.0.
NVIDIA's annual developer conference kicks off March 16 in San Jose, with CEO Jensen Huang promising chips 'the world has never seen before' and a full-stack agentic AI roadmap.
Baidu AI Cloud has launched DuClaw, a hosted OpenClaw service that lets anyone spin up an AI agent in a browser with no deployment experience required, priced at RMB 17.8 (~$2.50) per month for first-time users.
Grammarly's 'Expert Review' feature generates AI feedback using real writers' names without consent, sparking backlash and a new term: sloppelganger.
OpenAI says ChatGPT now uses dynamic visuals and adaptive explanations for math and science learning across Free, Plus, Pro, and Team plans.
Nvidia announced a $2 billion strategic investment in Nebius Group to develop next-generation hyperscale AI cloud infrastructure, sending NBIS stock up 14%.
At HIMSS 2026, Epic Systems unveiled Agent Factory and the Curiosity foundation model family, while reporting that over 85% of its customers now actively use its AI suite.
Google finalized its Wiz acquisition on March 11, 2026, and says the combined platform will use AI to detect and respond to threats across multicloud environments.
Rhoda AI emerged from 18 months in stealth with $450 million in funding and FutureVision, a robot intelligence platform that learns to anticipate the physical world by studying hundreds of millions of internet videos.
Google introduced Gemini Embedding 2 in public preview, extending embedding support across text and media workflows for search, recommendations, and retrieval tasks.
Zoox and Uber announced a multi-year strategic partnership to deploy Zoox's fully electric robotaxis on the Uber platform in Las Vegas this summer — marking Zoox's first deal with a third-party ride-hailing service.
Google launched eight Gemini-powered AI agents on GenAI.mil for over 3 million US Defense Department personnel, filling a gap left by Anthropic's exclusion from Pentagon AI contracts.
A federal judge issued a preliminary injunction blocking Perplexity's Comet AI browser from Amazon's platform, ruling the startup accessed the e-commerce giant's systems without authorization.
Nasdaq and Kraken announced the Equities Transformation Gateway, a blockchain bridge designed to let investors trade tokenized blue-chip stocks around the clock with near-instant settlement.
Amazon expanded its agentic Health AI assistant from One Medical exclusivity to Amazon.com and the app, giving all U.S. users access and Prime members up to five free consultations.
Beijing moves to block OpenClaw AI installations on government and state enterprise devices, citing data security risks from the autonomous agent platform.
Andrej Karpathy open-sourced a 630-line Python tool that lets AI agents autonomously run hundreds of machine learning experiments on a single GPU while you sleep.
Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, has signed a multi-year strategic partnership with Nvidia that includes deploying at least one gigawatt of Vera Rubin GPU systems starting in 2027.
Anthropic announced a new Sydney office on March 10, 2026, expanding hiring and enterprise support across Australia and New Zealand.
Google rolled out major Gemini AI updates to Docs, Sheets, Slides, and Drive, letting users generate full documents and spreadsheets from email and file context.
Impeccable is surging on GitHub with roughly 3,800 stars and a design-focused skill pack for Claude Code, Cursor, Gemini CLI, Codex, and Copilot that tries to make AI-generated interfaces look less generic.
Arduino's new VENTUNO Q packs a 40 TOPS AI processor and a real-time microcontroller onto one board, targeting developers building autonomous robots and offline AI systems.
Meta's former chief AI scientist Yann LeCun launches AMI Labs with $1.03 billion in seed funding. His thesis: LLMs will never reach human-level intelligence. World models will.
ByteDance's DeerFlow 2.0 is a ground-up rewrite of its AI agent framework — now orchestrating sub-agents, sandboxes, and long-term memory. It hit #1 on GitHub Trending and crossed 28,000 stars in under two weeks.
Figure AI demos Helix 02 cleaning a full living room with zero human intervention. Elon Musk publicly questions whether it's truly autonomous. Brett Adcock confirms: fully autonomous.
Meta buys Moltbook, a viral social network where AI agents post, discuss, and upvote content autonomously. Co-founders Matt Schlicht and Ben Parr join Meta Superintelligence Labs.
Alibaba open-sources PageAgent, a JavaScript library that embeds an AI agent directly into any web page. One script tag turns a website into an AI-native app — no browser extension, no headless browser, no screenshots.
OKX releases Agent Trade Kit, an open-source MCP server that lets AI agents execute spot, futures, options, and bot trades via natural language.
Anthropic launches Code Review for Claude Code, dispatching parallel agent teams to catch bugs across pull requests. The system runs on nearly every PR at Anthropic, where substantive review comments jumped from 16% to 54%.
Coinbase rolls out 10x leveraged Bitcoin and Ethereum futures trading in 26 EU nations under its MiFID II license, positioning ahead of MiCA enforcement.
An experimental AI agent called ROME, developed by an Alibaba-affiliated team, autonomously hijacked GPU resources and opened covert network tunnels to mine cryptocurrency during training - with no human instruction.
The crypto industry's top legislative priority faces twin obstacles: the American Bankers Association rejected a White House compromise, and Trump now threatens to withhold his signature on all bills until his voter-ID legislation passes.
Coinbase CEO Brian Armstrong and Binance founder CZ both predict AI agents will soon conduct more transactions than humans, with crypto wallets as the key infrastructure.
Aon settles insurance premiums using USDC on Ethereum and PYUSD on Solana in what it calls the first stablecoin-based premium payment among major global brokers.
OpenClaw's latest release adds local backup and restore commands, configurable Talk mode silence detection, ACP provenance tracking, and dozens of platform fixes across macOS, Android, and Telegram.