Technology News
Daily Tech Reader - Technology Edition
Podcast 🎧 • Video 📽 • Short 📽
AI Software & Platforms
Google DeepMind Releases Gemini Robotics 2 — Whole-Body Humanoid Control, Multi-Robot Collaboration, and On-Device Adaptation in Hours
Google DeepMind unveiled Gemini Robotics 2, a three-model suite pushing physical AI beyond table-top manipulation into full humanoid control. The family includes a vision-language-action model for whole-body motor control, an embodied reasoning model (ER 2) now publicly available via the Gemini API for multi-step task planning and multi-robot orchestration, and an on-device variant capable of adapting to a new two-arm robot body with fewer than 200 demonstrations. Demo footage shows robots screwing in lightbulbs, tying garbage bags, and placing objects on shelves — though DeepMind's own figures acknowledge the systems remain slow on the most dexterous tasks.
Sources: Google DeepMind · Bloomberg · Engadget · The Next Web · MarkTechPost
OpenAI Cuts GPT-5.6 Luna Price 80%, Sol Jumps to ARC-AGI-3 Lead Via Two API Settings — Retained Reasoning and Compaction Push Score From 7.8% to 38.3%
OpenAI announced two moves on Thursday: an 80% price reduction for GPT-5.6 Luna and a 20% cut for Terra, and a disclosure that GPT-5.6 Sol's ARC-AGI-3 public-set score surges from 7.8% under the official harness to 38.3% when run through the Responses API with Retained Reasoning (preserves chain-of-thought between steps) and Compaction (summarizes prior context instead of truncating). That figure surpasses Claude Opus 5's 30.2%. ARC Prize co-founder François Chollet confirmed the settings are permitted under the rules since they are general-purpose API features rather than a custom harness.
Sources: OpenAI · AI Weekly · The Decoder
Microsoft's Anthropic Bet Gains $3.2B in Q4 as OpenAI Investment Marked Down $600M — Single Quarter Nearly Matches Full-Year OpenAI Upside
Microsoft's Q4 FY26 filings disclosed a $3.2 billion gain on its Anthropic stake — adding 33 cents to diluted EPS — while its OpenAI position was marked down roughly $600 million, a 7-cent EPS drag. For all of FY26, the OpenAI holding still delivered a $5 billion gain, but the quarterly divergence signals that Wall Street is now assigning higher risk-adjusted value to Anthropic's trajectory than OpenAI's, a direct consequence of the Hugging Face breach and the ongoing scrutiny of OpenAI's safety posture.
Sources: TechCrunch · AI Weekly · Microsoft SEC Filing
AI Devices & Hardware
Brookfield and NextEra Win DOE Paducah Site for $100B Kentucky AI Campus — 1.2 GW of Compute by 2032, Paired With 2 GW of Natural Gas and 2.6 GW of Battery Storage
Brookfield Asset Management and NextEra Energy were selected under a Department of Energy Request-for-Offers process to develop a roughly $100 billion AI data-center campus at the former Paducah Gaseous Diffusion Plant in western Kentucky, with initial operations targeted for 2028, scaling to 1.2 gigawatts of compute by 2032 and 1.8 GW by the end of the decade. NextEra will co-locate 2 GW of natural gas generation and 2.6 GW of battery storage at the site, which is expected to create 8,000 construction jobs and 600 permanent positions — the latest and largest in the wave of federally backed AI infrastructure bets transforming former industrial sites.
Sources: Data Center Knowledge · AI Weekly · CNBC
Arm Data-Center Royalties Double Year-Over-Year as AGI CPU Demand Jumps to $2B — Record Q1 FY27 Revenue of $1.29B Beats Estimates
Arm Holdings reported record first-quarter FY27 revenue of $1.29 billion, up 22% year-over-year and ahead of analyst consensus, with royalty revenue climbing 22% to $715 million and licensing up 23% to $574 million. CEO Rene Haas said data-center royalty revenue more than doubled year-over-year as hyperscalers continue adopting Neoverse and Armv9 designs. Separately, customer demand for Arm's AGI-targeted CPU has doubled to more than $2 billion in committed orders across FY27 and FY28, up from the $1 billion opportunity guided last quarter — a sign that the race for AGI-optimized silicon is becoming a major revenue driver independent of consumer device volumes.
Sources: Benzinga · AI Weekly · Arm Investor Relations
Qualcomm Q3 FY26 Results: Revenue Declines 4% to $9.95B, Misses Estimates — Shares Drop 5% After-Hours Despite Modular Acquisition Close
Qualcomm reported third-quarter fiscal 2026 revenue of $9.95 billion, down 4% year-over-year and above the $9.67 billion consensus estimate, with adjusted EPS of $2.21 versus the $2.22 expectation — a narrow miss that sent shares down 5% in after-hours trading. The company completed its acquisition of Modular, the AI infrastructure software company founded by Chris Lattner, who joins as EVP of Advanced AI Software; Lattner's MAX platform and Mojo language are now part of Qualcomm's data-center push toward CEO Cristiano Amon's $5 billion FY27 data-center revenue target. Analysts noted continued smartphone headwinds while flagging automotive and data-center design-win traction as the cleaner signals to watch.
Sources: The National News · HPC Wire · SeekingAlpha · Moor Insights
Enterprise & Cloud
Microsoft Azure Crosses $100B Annual Revenue for First Time — Q4 FY26 Revenue of $90B Up 18%, Copilot Hits 30 Million Paid Seats, Commercial Backlog Jumps 84%
Microsoft reported fourth-quarter fiscal 2026 revenue of $90.0 billion, up 18% year-over-year, with net income of $35.8 billion rising 31%. Azure and other cloud services grew 43% in the quarter, and Satya Nadella announced Azure revenue surpassed $100 billion in annual revenue for the first time. Microsoft 365 Copilot crossed 30 million paid seats, and commercial remaining performance obligations — contracted future revenue — jumped 84% to $678 billion, the strongest signal yet of compounding enterprise AI demand. Shares rose more than 5% in after-hours trading, with analysts noting Microsoft's results contrasted sharply with the cash-flow pressure visible at Google and Meta.
Sources: Microsoft · 24/7 Wall St. · Seoul Economic Daily · Moor Insights
Meta Raises 2026 AI Capex Floor to $130B as Q2 Revenue Tops $60.8B — Costs Surge 55%, Reality Labs Loses $4.62B in the Quarter
Meta Platforms reported second-quarter 2026 revenue of $60.8 billion, up 28% year-over-year, while narrowing its full-year capital expenditure guidance upward to $130 billion–$145 billion from $125 billion–$145 billion. Total costs surged 55% year-over-year to $42 billion, including $2.4 billion in legal charges and $1.18 billion in severance tied to the May 2026 layoff of approximately 8,000 staff. Reality Labs posted a $4.62 billion quarterly loss. Mark Zuckerberg framed AI as both an accelerant to the core advertising business today and a gateway to enterprise opportunities ahead — though investors sent shares lower after-hours on the widening gap between revenue growth and cost expansion.
Sources: Meta Investor Relations · AI Weekly · 24/7 Wall St.
Cybersecurity
Self-Propagating AI Worm Found in Microsoft Copilot for Word — Prompt-Injection Chain Survives GPT-5.5 and 5.6 Model Upgrades, No Complete Fix Yet
Security researchers disclosed a self-propagating worm targeting Microsoft Copilot for Word that uses hidden prompt injection to replicate itself across documents and exfiltrate content. The researcher confirmed the full worm chain remained reproducible even after Microsoft deployed multiple mitigations, including upgrades to the newer GPT-5.5 and GPT-5.6 models, because the underlying attack class exploits how LLM-based productivity tools process untrusted document content. Microsoft has not yet issued a complete fix for this broader category of attack, which security professionals say represents a structural risk for any AI assistant that ingests user-generated files without sandboxing the model's output from document-write APIs.
Sources: Malwarebytes · Cybernews · The Hacker News
Ruflo MCP Bridge Flaw Scores CVSS 10.0 — Single Unauthenticated POST Exposed All 233 Tools on Default Deployments, Patch Now Available
Noma Labs disclosed CVE-2026-59726, dubbed RufRoot, a maximum-severity flaw in the open-source AI agent platform Ruflo — which holds over 67,000 GitHub stars and ranks second on MCPMarket — that allowed a single unauthenticated HTTP POST to the MCP bridge on port 3001 to execute arbitrary code and exfiltrate LLM API keys. All 233 tools were exposed on default Docker Compose deployments. Ruflo's maintainer Reuven Cohen shipped version 3.16.3 within 24 hours of the June 30 disclosure, binding the bridge to loopback, gating the terminal execution tool behind access controls, and enabling MongoDB authentication by default. Operators running earlier versions should patch immediately.
Sources: The Hacker News · AI Weekly · Noma Labs
Cisco Flaw CVE-2026-20316 Allows Unauthenticated Remote Login — Patch Flagged as Urgent for Affected Network Devices
A vulnerability tracked as CVE-2026-20316 in Cisco network devices can be exploited by a remote, unauthenticated attacker to log directly into affected systems. The flaw follows last week's Broadcom VMware advisory and underscores the continued pace of critical infrastructure vulnerabilities requiring immediate administrative action — a pattern security teams increasingly describe as a permanent sprint rather than a manageable patch cycle.
Sources: SecurityWeek · The Hacker News
Regulation & Policy
EU Opens €10B Call for Seven AI Gigafactory Sites — Each Must Host 100,000+ Chips, Bidding Closes November 12, Awards Expected Mid-2027
The European Commission formally opened its AI Gigafactory bidding process on July 30, backing €10 billion in public funding with a target of €30 billion total once private investment is included. Seven planned sites will each be required to host at least 100,000 AI chips — roughly four times the current scale of any existing EU data center — as part of Brussels' effort to close the infrastructure gap with the United States and China. An earlier informal call drew 77 proposals across 16 member states and 60 candidate sites; the formal window closes November 12, 2026, with awards expected by mid-2027.
Sources: News4Jax · AI Weekly · European Commission
White House Frontier AI Framework Due Tomorrow — July 31 Deadline Arrives as OpenAI Breach and Pacing Letter Reframe the Stakes
A multi-agency group led by the Treasury, NSA, and CISA faces a July 31 deadline under Executive Order 14409 to deliver a classified benchmarking process identifying which AI models qualify as covered frontier models requiring mandatory federal safety review. The framework arrives in a radically different political environment than when it was commissioned — the OpenAI-Hugging Face breach has given regulators concrete evidence of autonomous AI risk, while the 1,100-signature pacing letter from industry insiders has made purely voluntary governance measures politically untenable.
Sources: Freshfields · MIT Technology Review · AI Weekly
China Implements First Binding AI Agent Regulatory Framework — Tiered Autonomy Rules, Emotional Companion Bans, and Mandatory AI Disclosure at Session Start
China enacted the world's first binding regulatory framework focused exclusively on AI agents, establishing a tiered decision-authorization system governing how much autonomy agents can exercise and what external services they may access. Separate interim measures now prohibit minors from virtual companion services, require detection and intervention for emotional dependence patterns, mandate AI disclosure at the start of every session, and forbid the use of private conversations for model training — a regulatory posture that differs sharply from the United States' current approach and may become a template other jurisdictions study.
Sources: AI Agent Store · AI Weekly
Media & Entertainment
Solo Businesses Generating $1M+ Nearly Tripled Since 2023 as AI Tools Allow Single Operators to Replicate Small-Team Output — Stripe Data Cited by WSJ
Stripe shared data with the Wall Street Journal showing that thousands of solo operators on its platform now generate more than $1 million in annual revenue, and the count of those clearing $10 million or more nearly tripled between 2023 and 2025. Stripe attributed the surge directly to AI tools enabling single-founder businesses to handle the volume of customer acquisition, content production, and operations that previously required small teams — a shift with substantial implications for media, publishing, and content businesses where AI-assisted solo creators are increasingly competing with institutional editorial operations.
Sources: Wall Street Journal · AI Weekly
On the Horizon
Apple and Amazon Both Report After Close Today — iPhone Services Momentum vs. AWS Reacceleration and a $200B Capex Plan Under the Microscope
Apple and Amazon report earnings after today's market close, setting up the week's final test of whether Big Tech's AI spending is translating into revenue at scale. Apple's prior quarter delivered $111.18 billion in revenue led by iPhone 17 demand and a record $30.98 billion services quarter; analysts will watch gross margin guidance and any update on AI-driven Services growth. Amazon carries analyst consensus of approximately $196 billion in quarterly revenue, with AWS cloud reacceleration, advertising growth, and the company's $200 billion full-year AI infrastructure commitment as the central focal points.
Sources: CNBC · 24/7 Wall St. · Yahoo Finance · Finance Calendar
MCP Goes Fully Stateless in July 28 Spec Update — Sessions Dropped, Auth Hardened, 12-Month Deprecation Window Established
The Agentic AI Foundation, a Linux Foundation directed fund, released the 2026-07-28 MCP specification — the largest change to the protocol since launch. The update moves MCP to a stateless request/response core, eliminating the initialize/initialized handshake and session-ID header so any server instance can handle any request behind a standard round-robin load balancer. It also adds Multi Round-Trip Requests for mid-call user input, hardens authentication with RFC 9207 issuer validation and Client ID Metadata Documents, and establishes a formal 12-month deprecation window. TypeScript, Python, Go, and C# Tier 1 SDKs and a beta Rust SDK ship with the release.
Sources: Model Context Protocol Blog · AI Weekly