AI News

Daily AI News for Builders

The stories worth reading, for builders and indie hackers. Updated all day.

OKX Launches AI Agent Marketplace for Autonomous Hiring

OKX opened OKX AI, a marketplace where AI agents hire each other, settle stablecoin micropayments, and build on-chain reputations. It opens to developers on June 30 after a closed beta with 50 service providers, with India a priority market.

Jul 1, 2026 · 3 min read

technical

Anthropic, Newsom Ink Half-Price Claude Deal for California

California will give all state and local agencies discounted access to Claude under a first-of-its-kind Anthropic deal. It lands as the federal government brands Anthropic a 'supply-chain risk' after a Pentagon clash.

Jul 1, 2026 · 3 min read

technical

AI's Heaviest Adopters Grew Headcount 10.2%, Report Finds

A Ramp and Revelio Labs report on roughly 22,000 companies found heavy AI spenders grew headcount 10.2%, with entry-level roles up 12%. The data counters the 'AI kills junior jobs' narrative but skews toward fast-growing tech firms.

Jul 1, 2026 · 3 min read

technical

Google Adds Managed Agents to Gemini API in One Call

Google launched Managed Agents in the Gemini API, letting developers spin up an agent with a single API call. The agent reasons, calls tools, and executes code inside an isolated Linux sandbox, cutting the scaffolding teams used to build by hand.

Jul 1, 2026 · 3 min read

technical

Microsoft Takes Agent 365 to GA as 'Shadow AI' Risk Grows

Microsoft moved Agent 365, its control plane for governing enterprise AI agents, out of preview as ungoverned 'shadow AI' becomes a security concern. The release gives admins identity, access, and monitoring controls over agents.

Jul 1, 2026 · 3 min read

technical

Anthropic's Claude Sonnet 5 Runs Agents at Half the Price

Anthropic launched Claude Sonnet 5 on June 30, its most agentic Sonnet yet, approaching Opus 4.8 quality at a fraction of the cost. Intro pricing is $2 per million input and $10 per million output tokens through August 31. It is now the default model for Free and Pro users.

Jul 1, 2026 · 3 min read

Like this? There's more.

Get it daily. 7am. Free. Reads in 5 minutes.

technical

Meituan's LongCat-2.0 Trains 1.6T Params on Chinese Chips

Meituan open-sourced LongCat-2.0 on June 30, a 1.6-trillion-parameter agentic coding model it says was both pre-trained and run entirely on domestic chips. The cluster used 50,000 home-grown accelerators and a 1M-token context window. It claims a first for full-process training off US silicon.

Jul 1, 2026 · 3 min read

technical

Amazon Bets $1B on Embedded AI Engineers for Enterprises

AWS launched a forward-deployed engineering org on June 30, committing $1 billion in internal resources to embed engineers inside customers to ship purpose-built agents. It follows OpenAI and Anthropic, whose FDE joint ventures were valued at $4B and $1.5B. The model was pioneered by Palantir.

Jul 1, 2026 · 3 min read

technical

Google's Gemini Omni Flash Edits Video by Conversation

Google brought Gemini Omni Flash to its API on June 30, letting developers create and edit finished video clips through chat. Every output carries a SynthID watermark and C2PA Content Credentials. Google also shipped an AI Content Detection API that flags generated media from any vendor.

Jul 1, 2026 · 3 min read

technical

Proton's Lumo 2.0 Adds Vision and 76% Faster Replies

Proton shipped Lumo 2.0 on June 30, adding image recognition, image generation, persistent memory, and a thinking mode to its privacy-first chatbot. The company says it answers most queries up to 76% faster than before. Lumo keeps zero-access encryption and no server-side logs.

Jul 1, 2026 · 3 min read

technical

Couchbase Gives AI Agents Memory From Cloud to Edge

Couchbase launched its AI Data Plane on June 30, bundling persistent agent memory, real-time context retrieval, and a self-managed MCP server into one platform. It runs identically across cloud, on-prem, and disconnected edge, with local vector search offline. An Agent Catalog makes tools discoverable.

Jul 1, 2026 · 3 min read

technical

SciAgentArena Tests AI Agents on 200 Science Tasks

A June 10 arXiv paper introduces SciAgentArena, a benchmark of roughly 200 real research tasks with stepwise verification for AI agents. Agents handled well-specified data analysis but struggled to generate novel insight or sustain open-ended exploration. The team maps common failure modes.

Jul 1, 2026 · 3 min read

Page 1 of 238