SpaceXAI Launches Grok Bot — AI Agents as Teammates Grok is now an AI ’teammate’ you can assign work — The Verge SpaceXAI today introduced Grok Bot, a persistent AI agent service that operates like an independent “AI teammate.” Bots share a cloud computer environment and can sign into apps, tools, and websites on a user’s behalf to complete multi-step workplace tasks autonomously. SpaceXAI says it’s been using the system internally across engineering, sales, and finance, with results that include finding new accounts and pulling receipts from email.

🌍 Frontier Models & Milestones

ChatGPT and Gemini both just passed 1 billion users — The Verge The two dominant AI assistants hit the same staggering number on the same day. Google CEO Sundar Pichai confirmed Gemini now has 1 billion monthly active users — Google’s fastest product ever to reach the milestone — with 63% of users engaging through voice and the model generating over 150 million images daily. OpenAI’s ChatGPT crossed the same threshold earlier this year. The simultaneous milestone underscores how quickly consumer AI has gone mainstream.

His Start-Up’s Goal: AI That Is Trainable and Not Controlled by a Big Company — The New York Times Igor Babuschkin, who co-founded xAI with Elon Musk before departing a year ago, has unveiled River AI — a startup building a home computer that lets people own and run AI without oversight from any company, including River AI itself. Technologies will be open-sourced. Babuschkin envisions AI trained to benefit individuals rather than centralized interests.

OpenAI launches ChatGPT desktop app for Linux — TechCrunch. OpenAI extends its desktop client to Linux, the last major platform without native support.

As AI safety concerns mount, three pioneers make the case for staying open — TechCrunch. At Ai4, Geoffrey Hinton, Fei-Fei Li, and Andrew Ng debated regulation and open-source access as China advances.

📱 Made by Google 2026

Everything announced at Made by Google ‘26: Pixel 11, Pixel Watch 5, Pixel Tag, and tons of Gemini features — TechCrunch Google’s annual hardware event in New York delivered the full Pixel 11 lineup (11, 11 Pro, 11 Pro XL, 11 Pro Fold), all powered by the Tensor G6 chip. Highlights include 120x Pro Res Zoom on the 11 Pro, a new HiLight LED notification ring, and a starting 256GB base storage. Google also launched Pixel Tag — a direct AirTag rival — and spent much of the event on Gemini AI integrations baked deeply into every device.

Google’s Pixel Watch 5 dives deeper into AI and health — The Verge. The $399 Pixel Watch 5 gets a Qualcomm processor bump and a slightly larger battery; the real story is its expanded Gemini-powered health features and new ASL support for Live Transcribe.

💰 Funding & Market

AI coding startup Cognition reportedly already in talks to raise at $40B valuation — TechCrunch Cognition, maker of the autonomous AI software engineer Devin, is in early talks for a new round that would value it above $40 billion — less than three months after closing $1 billion at a $26 billion valuation. The company’s annualized revenue run rate is now approaching $1 billion, roughly double where it stood at the last raise, and 89% of all code committed internally is now written by Devin itself.

Lovable confirms new $13.3B valuation, raises another $400M — TechCrunch. The AI app-builder reached $500M in annualized run rate in June and now secures another large growth round.

OpenAI-backed Thrive Holdings raises $2B to bring AI to the enterprise — TechCrunch. Backed by SoftBank and D1 Capital, Thrive Holdings closes at a $12 billion valuation targeting enterprise AI deployment.

AI code-testing startup Blacksmith’s valuation jumps almost 10x in less than a year — TechCrunch. Revenue up 10x in a year; the firm now sits at $550M as AI-generated code drives demand for automated testing.

Nvidia’s Risky Business — Stratechery. Ben Thompson analyzes how Nvidia’s novel financing mechanisms — drawing on insurance floats and pension funds — significantly raise the risk profile behind its dominance.

🛠️ Developer Tools & Open Source

Modular 26.5: Mojo 1.0 is here! — Modular Three years after its debut, Mojo has reached 1.0 — locking in a stable, backward-compatible API and signaling production readiness. The language combines Python-like syntax with Rust-level memory safety and systems performance, designed specifically for heterogeneous AI hardware. The 26.5 release adds Python-style lambda closures, memory-safety diagnostics, and a Pointer type. Modular has teased open-sourcing the compiler itself at ModCon on August 18.

Paperclip (GitHub Repo) — DevOps / TLDR. Open-source Node.js server and React UI that orchestrates teams of AI agents (Claude Code, Codex, Cursor, Gemini) to run a business — assigning roles, budgets, reporting lines, and permissions.

Code-Graph-RAG (GitHub Repo) — DevOps / TLDR. Parses multi-language codebases with Tree-sitter, stores structure as a knowledge graph in Memgraph, and exposes an MCP server so Claude Code and other clients can query and edit code in plain English.

CNCF Announces Graduation of Cloud Native Buildpacks — CNCF. The project — which builds OCI-compliant containers directly from source code — graduates with 535 contributors across 164 organizations including Google, GitLab, and Bloomberg.

Kiro CLI: Cloud Sessions Preview and Smarter Command Menus — Kiro. Kiro’s CLI now lets you start a session with --cloud, disconnect, and resume from any machine while the agent keeps working.

Claude Code 2.1.228 — Anthropic. Bug fixes: interactive sessions that stopped redrawing, Git not found on Windows, and /tui reverting the active model.

🔒 Security & Policy

Cloudflare DDoS Threat Report H1 2026: 1 Tbps attacks soar as DNS floods and geopolitical tensions drive a new wave — Cloudflare Cloudflare mitigated 29.64 trillion HTTP DDoS requests in the first half of 2026. The starkest finding: network-layer attacks exceeding 1 terabit per second rose more than six-fold from Q1 to Q2. April was the peak month at 165 petabytes of volume. The decline following Operation PowerOFF — a 21-country action targeting 75,000+ DDoS-for-hire users — offers a rare data point on the impact of international law enforcement coordination.

Everything hackable will get hacked — Vercel. A sober assessment: defenders currently have an edge using stronger closed models, but the gap with open-weight models is closing fast. The recommendation: use frontier models now to continuously improve your vulnerability discovery and remediation cycle.

EU AI Act high-risk provisions now enforceable — AI Agent Store. As of August 2, 2026, the EU AI Act’s high-risk requirements — risk management, human oversight, conformity assessment — are live, alongside transparency mandates requiring chatbots to identify as AI. Non-compliance risks fines up to €15M or 3% of global revenue.

Twitch streamers can now opt out from training Amazon’s AI — The Verge. Streamers can now exclude their VODs, clips, chat, and channel content from future Amazon generative AI model training.

Generated by claude-sonnet-4-6 · 2026-08-12T10:00:00Z