Technology News
Daily Tech Reader - Technology Edition
Podcast 🎧 • Video 📽 • Short 📽
AI Software & Platforms
Hassabis Steps Back — Kavukcuoglu Takes DeepMind CEO as Demis Moves to Alphabet Chief Scientist Role Focused on AGI Strategy
Demis Hassabis is stepping down as Google DeepMind's day-to-day CEO to become chair of DeepMind and Alphabet's newly created Chief Scientist, while longtime CTO Koray Kavukcuoglu takes over as SVP running Gemini, frontier research, and app teams.
Sources: Axios · AI Weekly
Jeff Dean Quits Google After 27 Years to Co-Found Discovery Loop — AI-for-Science Startup Aims to Automate the Full Experimental Method
Jeff Dean is leaving Alphabet alongside DeepMind VP Oriol Vinyals and Google Brain co-founder Quoc Le to launch Discovery Loop, a public benefit corporation that will begin by automating ML research before expanding into drug discovery, chip design, and biology.
Sources: Unite.AI · AI Weekly
Meta Launches Muse Code Terminal Agent on New Muse Spark 1.2 Model — Priced at $1.25/M Input Tokens, Available on macOS and Linux
Meta released Muse Code, a terminal coding agent built on its new coding-focused Muse Spark 1.2 model, with async background workers, local event logging, and bundled skills including /plan and /grill.
Sources: Meta AI Research · AI Weekly
Ninth Circuit Overturns Amazon Block on Perplexity's Comet Shopping Agent — Court Finds Users, Not Perplexity, Are the Ones Who "Access" Amazon
The Ninth Circuit reversed a lower-court order barring Perplexity's Comet agent from Amazon.com, marking the first federal appeals ruling on whether autonomous AI agents can legally act on behalf of users across the web.
Sources: Bloomberg Law · AI Weekly
Cloudflare Open-Sources Its Internal AI Agent Workspace — Cloudflare OS Bundles Browser Sessions, Code Runtime, and a Zero-Permission Agent Model
Cloudflare released the source code for Cloudflare OS, its internal agent workspace, including a Gatekeeper service that issues typed capability bindings and data-observation tracking that blocks any output leaking inaccessible resources.
Sources: Cloudflare Blog · AI Weekly
AI Devices & Hardware
SpaceX Goes All-In on Nvidia Vera Rubin — Musk Plans 2 GW Installed by Year-End and 10 GW by End of 2027, Starmind Satellite to Follow
On SpaceX's Q2 earnings call, Elon Musk announced the company will build its AI infrastructure exclusively on Nvidia's Vera Rubin architecture, with the first Starmind satellite slated to fly next year running an optimized Vera Rubin NVL72 in orbit.
Sources: American Bazaar Online · AI Weekly
AMD Q2 Revenue Jumps 50% to Record $11.5B — Data Center More Than Doubles to $6.7B but Shares Fall 7% on Underwhelming Q3 Outlook
AMD posted record quarterly revenue driven by EPYC and Instinct GPU demand, with Data Center now at 58% of total revenue, but the Q3 guide of roughly $13B disappointed investors already pricing in an AI-fueled rally.
Sources: AMD IR · AI Weekly
CoreWeave Books 360 MW in Indonesia in First Asia-Pacific Push — Three Data Centers Expected Online in 2028, Targeting Southeast Asian Government and Enterprise Demand
CoreWeave announced its expansion into Indonesia with three owned-and-operated facilities totaling 360 MW of contracted IT power, extending its 49-site global footprint to meet regional AI compute demand.
Sources: CoreWeave · AI Weekly
Cybersecurity
CISA Adds Three Flaws to KEV Catalog — Langflow RCE, Apache Tomcat Encryption Bypass, and N-able Auth Bypass All Under Active Exploitation
CISA added CVE-2026-9198 (Langflow remote code execution, CVSS 9.8), CVE-2026-34486 (Apache Tomcat encryption bypass), and the previously flagged N-able N-central authentication bypass to its Known Exploited Vulnerabilities catalog on August 5.
Sources: The Hacker News · CISA
npm Credential-Stealing Worm Spreads From Keyv to 868 Packages Across 1,381 Versions — Preinstall Script Targets Developer and CI Environments
A malicious worm first detected in keyv@6.0.0 spread via a preinstall script to steal credentials inside developer and CI pipelines, with monitoring firms reporting footprints ranging from 353 to 868 poisoned packages as of August 4.
Sources: The Hacker News · SafeDep · Aikido
Russian Hackers Abusing Hotel Wi-Fi Networks to Steal Microsoft 365 Credentials and Deploy Malware
Threat researchers disclosed an active campaign in which Russian-linked actors are exploiting hotel Wi-Fi networks as an initial access vector to harvest Microsoft 365 credentials and deploy malware against guests' devices.
Sources: Help Net Security
AI Developers Targeted via Trojanized GitHub Repositories — Attackers Embedding Malicious Payloads in Widely Forked ML Tooling Projects
Security researchers identified a supply-chain campaign targeting AI developers through trojanized GitHub repositories, with attackers embedding malicious payloads inside popular machine-learning tooling projects to compromise developer environments.
Sources: Help Net Security
Enterprise & Cloud
Microsoft Revenue Filing Shows $24B of Its $37B AI Run-Rate Comes From Reselling OpenAI — Internal Copilot Growth Remains Modest
A Bloomberg review of Microsoft's latest filing reveals the bulk of the AI revenue Satya Nadella touted at end-March is OpenAI API resale, not in-house Copilot product growth, raising questions about the durability of Microsoft's AI premium.
Sources: Bloomberg · AI Weekly
Microsoft Tells Engineers to Stop "Tokenmaxxing" — Division-Level AI Token Budget Targets Set, GPT-5.6 Becomes Default Internal Model
Microsoft EVP Jay Parikh issued a memo capping engineer AI spend after finding many teams were burning thousands of dollars monthly in tokens, with OpenAI's GPT-5.6 designated the new internal default for its lower cost.
Sources: 404 Media · AI Weekly
Anaconda Acquires AI Security Firm Enkrypt AI — Adds Red-Teaming Across 300+ Attack Categories and MCP Server Vulnerability Scanning to Platform
Anaconda folded Enkrypt AI's pre-deployment red-teaming, runtime guardrails, and EU AI Act compliance automation into the Anaconda Platform, citing Enkrypt's finding of 143,000 vulnerabilities across 73% of scanned MCP servers.
Sources: Anaconda · AI Weekly
Zoox Starts Paid Robotaxi Rides in Las Vegas on August 10 — Fares Include Base Charge Plus Distance and Time, Surcharges for Airport and Arena Trips
Zoox is ending its free-ride era following last week's NHTSA commercial exemption, launching paid service in Las Vegas while free testing continues in San Francisco and Austin pending state permits.
Sources: TechCrunch · AI Weekly
Regulation & Policy
Rust Bans LLM-Written Pull Requests — Maintainers Allow LLM Analysis and Review but Not Code Creation, Authors Must Disclose AI Use
The rust-lang/rust project published a formal LLM policy on August 5 barring AI-generated code from PRs, citing reviewer bandwidth and the loss of craftsmanship as a quality signal, with non-compliant PRs subject to closure without explanation.
Sources: Rust Blog · AI Weekly
Google to Remove Assistant From Android on September 4 — Gemini Becomes Sole Assistant on Android and Wear OS, No Option to Switch Back
Google confirmed Google Assistant will be phased out beginning September 4, 2026 across Android, Wear OS, paired headphones, and Android Auto, with Gemini taking over as the permanent replacement.
Sources: 9to5Google · AI Weekly
On the Horizon
The Google That Built AI Is Now Reorganizing Around It — Hassabis to Chair, Dean Out, Kavukcuoglu In, All in One News Cycle
Wednesday delivered the biggest Google leadership reshuffle in years: Hassabis moving to Alphabet Chief Scientist, Jeff Dean departing after 27 years, and Kavukcuoglu stepping up — all as Alphabet shares fell more than 5% on the combined announcements.
Sources: Axios · Unite.AI · AI Weekly
African Developers Choosing Chinese Open-Source AI Over US Closed Models — DeepSeek, Qwen, and Kimi Winning on Cost and Customization
A New York Times feature reports African developers are overwhelmingly adopting Chinese open-weight models for their lower cost and easier customization for local languages, with Uganda's Sunflower LLM — covering 31 languages and built on Qwen 3 — as a leading example.
Sources: New York Times · AI Weekly