This week’s Spotlight is: Claude as the New Browser
Netscape organized the Internet by becoming the gateway to every website. The App Store organized mobile. Claude is becoming the gateway to how work gets done, and it’s happening faster than either of those shifts did.
We are moving from systems of record and intelligence to systems of work. I am calling this layer Work-as-a-Service, or WaaS. The applications stay where they are. The orchestration moves up to the agent. The interface becomes intent.
But the browser analogy only takes you so far. Browsers were passive. You navigated them. Claude is active. It suggests, decides, and executes on your behalf. That is a far bigger shift than the browser delivered.
For founders in the SaaS era, you won by getting users into your product. In the Claude era, you win by being the thing the agent calls.
My advice for founders:
Design for delegation, not interaction. Ask what high-value task a customer would fully delegate to an agent if it actually worked. Once you have the task, assume the AI is your primary user, and humans are supervisors. That means API-first architecture, structured inputs and outputs, deterministic behavior wherever possible, and observable execution.
Make your data compound, not accumulate. Static data commoditizes. Compounding data builds defensibility. Every interaction with your product should generate a proprietary signal that sharpens the next one.
Treat context as a product. Context is the structured understanding of how your customer actually operates. Aggregate context across the tools your customer already uses, structure it cleanly, and maintain memory over time.
The most valuable company of the next decade may not be the best app. It may be whatever sits between you and all your apps, deciding what gets done. That kind of company wins by becoming the system the agent reaches for first.
Browsers unbundled websites. Claude is going to unbundle applications.
This week’s signals make the WaaS shift feel concrete, with Anthropic expanding Claude into critical infrastructure, Microsoft launching Scout as an always-on agentic assistant for Microsoft 365, and Uber capping Claude Code and other AI tools after internal demand outpaced budgets. Agents are becoming important enough to need partners, controls, procurement discipline, and deep workflow integration, which is exactly what happens when a new work interface starts moving from the edge into the enterprise core.
Full Weekend Edition below. 👇
Signals Shaping the Future of AI:
Infrastructure
Alphabet plans to raise $80 billion to fund AI infrastructure and global compute expansion. The financing plan is intended to support data center buildouts, cloud capacity, and AI infrastructure investments as Alphabet scales compute for AI products and services. Click here
Dow showcases thermal-management materials for AI-driven computing at COMPUTEX 2026. The release covers cooling materials, semiconductor packaging components, and reliability technologies designed for high-density AI server and rack environments. Click here
Intel lays out a full-stack AI infrastructure strategy spanning PCs, edge AI, foundational data centers, and intelligence centers. The roadmap spans PCs, edge AI, foundational data centers, and intelligence centers, covering Intel’s product positioning across client, edge, and data-center segments. Click here
NVIDIA and Marvell frame optical interconnects as a next-wave constraint for autonomous AI agents and distributed data centers. The discussion focused on high-bandwidth connectivity, optical data movement, and networking technologies needed to support AI systems across larger and more distributed compute environments. Click here
HPE raises its long-term targets after AI compute demand drives a sharp acceleration in revenue. The company reported accelerated revenue growth across servers, networking, and cloud-AI products tied to AI compute demand, and raised long-term guidance. Click here
NVIDIA’s Vera CPU expands its data-center roadmap beyond GPUs and increases competitive pressure on Intel and AMD. The launch positions NVIDIA to capture a larger share of the AI server stack by pairing CPUs, GPUs, networking, and software into a more integrated infrastructure platform for AI workloads. Click here
Enterprise
OpenAI’s Codex adoption expands beyond software developers into broader office and knowledge-worker workflows. The report describes Codex being used for tasks such as analysis, document work, scheduling, and cross-application workflows outside traditional software engineering. Click here
Uber caps employee use of Claude Code and other AI tools after internal spending exceeded budget expectations. The company is adding usage limits and cost controls after employee demand for AI coding and productivity tools outpaced its internal budget. Click here
Anthropic formalizes its Claude Partner Network with a services track, partner hub, and enterprise certification requirements. The program gives partners a structured way to support Claude deployments through services, certifications, resources, and customer implementation support. Click here
Meta launches an enterprise-focused AI business agent to automate daily operations across customer communication, sales, and service workflows. The agent is designed to work across Meta’s messaging platforms and business systems to support customer service, sales, bookings, and operational tasks. Click here
Microsoft launches Scout, an OpenClaw-inspired agentic assistant for the Microsoft 365 ecosystem. The product is designed to operate across Microsoft 365, with memory, skills, cloud desktop access, and controls for enterprise use. Click here
Capital Flows
Anthropic confidentially files for an IPO after raising $65 billion at a near-trillion-dollar valuation. The filing follows Anthropic’s large private financing round and begins the process for a potential public listing. Click here
SpaceX sets an IPO price of $135 per share, targeting a $1.75 trillion valuation and a record $75 billion raise. The offering would become the largest IPO in history and rank SpaceX among the most valuable publicly traded companies in the world. Click here
DeepSeek is reportedly raising $7 billion in its first external funding round at a valuation of up to $59 billion. Investors, including Tencent and CATL, are expected to participate as DeepSeek expands beyond chatbots into more compute-intensive agentic AI systems. Click here
NVIDIA acquires Kumo AI for more than $400 million to expand its enterprise graph-learning and predictive AI capabilities. The acquisition adds Kumo’s graph neural network technology to NVIDIA’s enterprise AI stack, strengthening its ability to support customer data modeling, forecasting, recommendations, fraud detection, and other structured-data workflows. Click here
Supabase raises $500 million at a $10.5 billion valuation as demand for AI-native development infrastructure accelerates. The company, whose platform powers many applications built with tools like Claude Code and Codex, has become a key backend provider for the growing wave of AI-assisted software development. Click here
Ramp raises $750 million at a $44 billion valuation as it expands AI-powered financial operations. The company is embedding AI agents across procurement, expense management, accounting, and budgeting workflows while building tools to help enterprises manage AI spending and token usage. Click here
Research
Microsoft Defender adds preview capabilities to detect, block, and investigate threats to AI agents. The release includes tools for monitoring agent behavior and identifying pros. Click here
Apica releases enterprise research showing that agentic AI could trigger a 9.5x surge in telemetry data. The research estimates how agentic AI adoption could increase observability data volumes and affect monitoring, telemetry pipelines, and infrastructure requirements. Click here
Microsoft launches ASSERT to turn natural-language policies into application-specific AI behavior tests. The framework lets developers convert written behavior requirements into tests for AI applications, agents, tool use, and policy adherence. Click here
Researchers publish BigFinanceBench to evaluate financial-research agents on workflow-grounded tasks. The benchmark tests financial research agents on tasks involving retrieval, reasoning, tool use, and decision support in finance workflows. Click here
Researchers publish EvoDS, a self-evolving autonomous data-science agent with skill learning and context management. The paper presents an autonomous data-science agent designed to learn reusable skills, manage context, and improve across data-analysis tasks. Click here
Policy
President Trump signs an executive order creating a voluntary federal review process for advanced AI models before public release. The order gives the federal government up to 30 days to review certain advanced AI systems for potential national security risks before deployment. Click here
The U.K. forces Google to give publishers an opt-out from AI Search aggregation and training-related exposure. The change gives publishers a way to avoid having their content included in certain AI Search experiences and related content-processing workflows. Click here
The White House confirms a June 2 executive order creating a voluntary frontier-model review pathway for national-security risks. The order establishes a voluntary federal process for reviewing advanced AI models before public release while avoiding mandatory preclearance. Click hereClick here
Global AI Strategy
The European Union launches a tech-sovereignty package to reduce dependence on foreign AI, cloud, chip, and data-center infrastructure. The package includes measures focused on domestic cloud resilience, AI infrastructure, semiconductor capacity, and expanded European data-center development. Click here
Canada unveils an “AI for All” strategy with a C$500 million tech growth fund and additional sovereign compute commitments. The plan targets domestic AI-company financing, Canadian-controlled compute, AI adoption, healthcare AI, and AI safety capacity as part of a broader sovereign AI industry push. Click here
Anthropic expands Claude Mythos access to critical-infrastructure organizations across more than 15 allied countries. The expansion gives organizations in sectors such as power, water, healthcare, communications, hardware, NATO, and allied national systems access to Anthropic’s cyber-focused AI model. Click here
Google breaks ground on a new AI-ready data center in Horndal, Sweden. The project expands Google’s European data-center footprint and includes a heat-recovery design for local energy reuse. Click here
Talent Signals
Each week, we spotlight key roles tied to the themes shaping this week’s AI headlines, connecting talent to the companies driving the news.
Sekai builds the AI-native workspace for teams operating across content, workflows, and digital experiences, embedding agents directly into the productivity layer rather than bolting them on as assistants. As enterprise software shifts toward agent-driven execution and headless interfaces, Sekai is hiring across engineering, product, and go-to-market. Open roles are listed on its careers page. Click here
Dust builds the multiplayer AI workspace that lets teams share, version, and govern agents across an entire organization, with permissions, memory, analytics, and tool integrations. Dust raised $40 million to expand its multiplayer-AI platform for the enterprise and is hiring across engineering, product, and customer engineering. Open roles are listed on its careers page. Click here
OpenRouter runs the multi-model exchange routing AI workloads at over 25 trillion tokens per week, with built-in failover, governance, and cost optimization for teams building applications on top of Claude, Gemini, GPT, and open models. OpenRouter raised $113 million at a $1.3 billion valuation and is hiring across systems, applied AI, and customer engineering. Open roles are listed on its careers page. Click here
You can see all the opportunities at Mayfield-backed AI companies here.
Social Signals
The most important conversations in AI are unfolding across social media, where top voices are shaping the next wave of signals and strategy. Here are some of the top social signals and their takes from the past week.
Elad Gil (Click here) — “The events of the last 6 months in technology are arguably among the most important in human history. The tools now increasingly exist for recursive self-improvement of models and agents. We are likely in very early lift off and exponential, largely unnoticed outside of tech.” Gil argues that recent advances in AI are not simply incremental improvements, but the early stages of systems that can help improve future generations of themselves. His post reflects a growing view among some technology leaders that the pace of AI progress is accelerating faster than broader public awareness or understanding.
Marina Favaro (Click here) — “The Anthropic Institute published ‘When AI builds itself,’ which shows Claude is accelerating AI development. Anthropic engineers, on average, ship 8x as much code per quarter as they did from 2021 to 2025. On open-ended coding problems where answers are unclear, Claude’s success rate is now 76%, a 50-point jump in just 6 months.” Favaro highlights a new analysis from Anthropic examining the path toward recursive self-improvement, in which AI systems increasingly contribute to the development of future AI systems. While noting that fully autonomous AI development is not here today, the report argues that current trends are making faster AI-driven progress more plausible, raising both opportunities and important questions around oversight, verification, and control.
Chamath Palihapitiya (Click here) — “This is increasingly true of the frontier models across a variety of evals. There is no single best model. At the top of the leaderboard, Opus 4.7, GPT-5.5, and Sonnet 4.6 appear almost indistinguishable, separated by less than 0.3 percentage points overall.” Reacting to new benchmark results, Chamath points to a growing convergence among leading frontier models. As raw model performance becomes harder to differentiate, attention increasingly shifts toward distribution, workflow integration, data advantages, and how intelligence is packaged into products.
