Google I/O Builds Gemini Intelligence Layer.
TL;DR
- Google I/O showcases Gemini 3.5 Flash: Google positions Gemini as a universal intelligence layer, integrating across devices and cloud, with Spark and Omni models signaling a new era of always-on, multimodal AI.
- Andrej Karpathy shifts to Anthropic: The departure of a key OpenAI co-founder and AI architect to a direct competitor significantly alters the industry's talent landscape and perceived strategic alignment.
- Anthropic acquires Stainless, cuts off SDK access: A strategic infrastructure acquisition by Anthropic effectively disarms OpenAI and Google by eliminating a critical developer tooling dependency, highlighting competitive leverage.
- Anthropic challenges Pentagon's supply-chain blacklisting: A federal judge calls the DOD's designation "spectacular overreach," setting a precedent for the legal and ethical boundaries of AI deployment in national security contexts.
- Nvidia's Q1 FY27 earnings loom: Tomorrow's report is critical for market sentiment, with focus shifting from revenue beats to forward guidance on sustained demand and hyperscaler capital expenditure.
Lead Story: Google I/O Builds Gemini Intelligence Layer
Google's annual developer conference delivered a clear statement: Gemini is now positioned as a universal intelligence layer. The keynote introduced Gemini 3.5 Flash, available today across products and APIs. CEO Sundar Pichai highlighted its 289 tokens per second, emphasizing a fourfold speed advantage for complex agentic tasks. The more powerful Gemini 3.5 Pro, already in internal use, will reach developers within the next month.
The product slate underscores a strategic shift toward pervasive AI. Gemini Spark, Google's always-on personal AI agent running on dedicated cloud VMs, directly challenges Claude Cowork's market position. DeepMind's Gemini Omni Flash, a multimodal video generation model, is now available to subscribers. Beyond models, Antigravity 2.0 offers 12x faster IDE performance as a standalone desktop app, while Android Halo integrates a dedicated AI agent hub into Android 17. Google also announced its most significant Search overhaul in 25 years, featuring persistent "information agents" for push alerts, complemented by a new $100/month AI Ultra tier.
Hardware announcements, carried over from last week's Android Show, included Googlebooks from several OEMs and Android XR smart glasses developed with Samsung. In a calculated counter, Apple scheduled its WWDC 2026 keynote for June 8 on I/O day, signaling its primary AI rival is Google, not OpenAI.
In Other News
Andrej Karpathy defects from OpenAI's orbit to Anthropic. The former OpenAI co-founder, Tesla AI director, and pioneer of "vibe coding" confirmed his move to Anthropic on Tuesday. He will join Nick Joseph's pre-training team, focusing on leveraging Claude to accelerate foundational research. This shift follows closely on the Musk v. Altman trial, where Karpathy's name surfaced repeatedly, and aligns with other key departures from OpenAI, including John Schulman and xAI's Ross Nordeen. As Gizmodo observed, this signals a gravitational pull of serious AI talent towards Anthropic, or at least away from Altman's sphere.
Anthropic acquires Stainless for ~$300M+, cuts off OpenAI and Google's SDK pipeline. Anthropic announced the acquisition of Stainless, a startup founded by ex-Stripe engineer Alex Rattray, and concurrently stated it will discontinue all hosted products. This includes the critical SDK generator relied upon by OpenAI, Google, Cloudflare, and Runway. Forbes noted that OpenAI had migrated to Stainless after its internal tooling proved insufficient, confirming Anthropic has strategically neutralized a rival's silent dependency. This marks Anthropic's second dev-infra acquisition, following Bun in December 2025, underscoring a deliberate strategy to control core agent connectivity infrastructure.
Anthropic and the Pentagon clash in DC appeals court. A three-judge panel heard oral arguments Tuesday in Anthropic's lawsuit challenging its blacklisting by Defense Secretary Pete Hegseth, who designated the company a supply-chain risk after it refused to grant the military unfettered access to Claude. Judge Karen Henderson was direct: "For the life of me, I do not see any evidence of maliciousness. To me this is just a spectacular overreach by the department." The DOJ countered that Anthropic's safety limits constitute an "untenable national-security risk." A written ruling is pending.
Meta restructures 7,000 employees into AI divisions ahead of layoffs. An internal memo obtained by Reuters outlined the formation of three new divisions: Applied AI Engineering, Agent Transformation Accelerator, and Central Analytics. These units are tasked with developing AI agents to automate existing human roles. Meta also replaced Llama with Muse Spark in its consumer products following Llama 4's underperformance relative to DeepSeek V3.
X / Social Pulse
Karpathy's concise X post — "I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative" — garnered nearly 3 million views within its first hour. The responses were sharply divided, interpreting the move either as a direct vote of no confidence in OpenAI or as a shrewd career pivot timed with the Musk verdict. Developer communities were consumed by assessing the engineering overhead OpenAI and Google will face in replicating Stainless's core functionalities. Gemini Spark's "runs on cloud VMs 24/7" feature immediately ignited comparisons to Claude Cowork, intensifying the agent-market competition. Notably, Nintendo's 6.8% stock surge, attributed by some to "AI fatigue" among investors, became an unexpected daily meme.
One to Watch
Nvidia's Q1 FY27 earnings report tomorrow after close. Consensus forecasts revenue between $78-78.5B and EPS of $1.77, reflecting 78% year-over-year growth. Early Vera Rubin chip samples are now shipping, promising a 10x inference cost reduction over Blackwell. Morgan Stanley elevated its conviction, describing a beat-and-raise scenario fueled by hyperscaler commitments of $725B in 2026 capex. The critical metric will not be the immediate quarter's performance, but whether forward guidance signals sustained demand or the onset of a market plateau.
Quick Hits
- Anthropic revenue run rate hit $30B+ annualized — a significant jump from $9B at the close of 2025, solidifying its position as CNBC Disruptor 50 No. 1 with a valuation now reportedly exceeding OpenAI's.
- AI venture funding reached $255.5B in Q1 2026 — surpassing the entirety of 2025 within a single quarter, with AI capturing 80% of all global VC per PitchBook.
- Mistral AI acquired Emmi AI for ~300M euros — a physics simulation specialist, as Mistral develops a European Mythos alternative for banks impacted by Anthropic's restrictions.
- Samsung and its union continue intense negotiations in 14+ hour talks, though no deal yet; a court granted a partial injunction requiring 7,087 essential workers to report if over 45,000 walk out Thursday.
- Hitachi partners with Anthropic to integrate infrastructure AI across 290,000 employees, aiming for 100,000 AI-proficient workers and covering critical sectors like rail, power grids, manufacturing, and cybersecurity.
- EU AI Act omnibus extends high-risk compliance deadlines by 16 months to December 2027; a ban on "nudifier AI" is now effective December 2026 per Latham & Watkins.
Tuesday's events delivered a trifecta of market-defining shifts. Google I/O confirmed Gemini 3.5's tangible arrival and its ambition as a pervasive intelligence layer. Karpathy's strategic talent migration to Anthropic fundamentally redrew the competitive map. Concurrently, Anthropic's Stainless acquisition militarized the infrastructure layer, denying critical tooling to rivals. Coupled with a federal judge dismissing the Pentagon's blacklisting as "spectacular overreach," the message is clear: the AI race now extends far beyond model performance. It encompasses talent acquisition, strategic tooling, legal standing, and control over the connective architecture that integrates AI into the global economy.
Sources
- CNET — Google I/O 2026 live updates
- Wired — Live blog
- Yahoo Tech — Everything announced
- Android Authority — I/O 2026
- Reuters — Karpathy joins Anthropic
- CNBC — Karpathy hire
- TechCrunch — Karpathy pre-training team
- Axios — Karpathy
- Business Insider — Karpathy
- Gizmodo — Karpathy
- Anthropic — Stainless acquisition
- TechCrunch — Stainless
- Forbes — Stainless SDK access
- AI Insider — Stainless $300M+
- CNBC — Anthropic vs DOD court
- Axios — Court arguments
- CNBC — Disruptor 50 No. 1
- Reuters — Meta restructuring
- Reuters — Mistral acquires Emmi AI
- Kiplinger — Nvidia earnings preview
- CNBC — Morgan Stanley on Nvidia
- Reuters — Samsung talks narrow differences
- Reuters — Samsung strike explainer
- PitchBook via Crowdfund Insider — Q1 AI VC funding
- Latham & Watkins — EU AI Act
- Nippon.com — Hitachi-Anthropic partnership
- Investing.com — Hitachi-Anthropic
- Glitchwire — Gemini 3.5 speed
- Let's Data Science — Google Search AI
Lock in. M. mazen@thorterminal.com