💰 Anthropic Raises $65 Billion for IPO

Anthropic has closed a $65 billion Series H round at a $965 billion post-money valuation, marking what could be its final private fundraise before a public debut.

Of the $65 billion, $15 billion consists of previously committed hyperscaler investments, including $5 billion from Amazon. The company's revenue run rate crossed $47 billion this month, with a 130% revenue surge projected to deliver its first operating profit.

Key Insights:

• $965 billion post-money valuation - Series H co-led by Altimeter Capital, Sequoia, and Greenoaks alongside institutional investors including Blackstone and Fidelity

• $15 billion from hyperscalers - A portion consolidates previously committed investments from cloud partners, including $5 billion from Amazon announced in April

• Revenue run rate at $47 billion - Crossed this month, driven by accelerating enterprise adoption of Claude Code

• 130% revenue surge projected - The Wall Street Journal reported this would bring Anthropic to its first operating profit

• Claude Opus 4.8 launched same day - Released with improved agentic capabilities, coding performance, and a focus on honesty and self-correction

The Bigger Picture: The $965 billion valuation tells you something about where the market thinks AI is heading: this is no longer venture speculation, it is infrastructure pricing. Anthropic is being valued like a utility, not a startup. The IPO signal matters as much as the round size. When a company raises at near-trillion-dollar valuations, public markets are the next logical step, changing how enterprises evaluate AI vendors. Companies already building on Claude are not betting on a model, they are betting on a publicly accountable business with institutional oversight.

Read More →

📈 Robinhood Opens Trading to AI Agents

Robinhood now lets users create a dedicated account for an AI agent and fund it with real money, giving the agent authority to buy and sell stocks autonomously. Users link AI agents through the Model Context Protocol, an open standard connecting AI systems to apps and data. The company provides push notifications for every agent trade, with the ability to pause AI trading at any time. Robinhood warns explicitly that agentic trading involves significant risk and takes no responsibility for agent-generated losses.

Key Insights:

• Dedicated agent accounts - Users create a separate trading account for their AI agent, isolating risk from their main portfolio while granting autonomous trading authority

• MCP protocol powers the integration - The Model Context Protocol connects AI systems to Robinhood's platform, enabling standardized agent access across different AI tools

• Real-time transparency controls - Push notifications for every agent trade, a live activity feed, and the ability to pause AI trading instantly

• Equities first, more assets coming - Beta launches with stock trading, expanding to options, cryptocurrency, event contracts, and futures

• Gold Card AI agent included - Gold Card customers receive access to an AI-powered agent connected to a virtual card, extending autonomous AI capabilities beyond trading

The Bigger Picture: This is not about making trading easier. It is about removing humans from the decision loop on a mainstream consumer platform. Robinhood democratized finance; now it is doing the same for autonomous AI action. The Model Context Protocol at the center of this rollout is becoming the standard for how AI systems connect to real-world apps. If MCP becomes the dominant interface between agents and digital services, whoever builds the best MCP-native tools gains the widest reach. Robinhood just removed the technical complexity standing between retail investors and an autonomous AI portfolio manager.

Read More →

🏗️ Agentic AI Exposes a Structural Gap

85% of organizations want to be fully agentic within three years. 76% say their current operations cannot support that change. The gap, per MIT Technology Review, comes from layering AI agents onto existing human-designed workflows rather than rebuilding the underlying operating model. PwC UK's Chief AI Officer compares the approach to adding sticky tape to parts of a structure that is already breaking.

Key Insights:

• 85% aim to go agentic - A large majority of organizations want full agentic capability within three years, but most lack a credible path to get there

• 76% lack the infrastructure - Fewer than one in four report having the systems and workflows needed to support an agentic operating model today

• 30%-50% process acceleration on offer - Early deployments in customer service, HR, and sales show AI agents can accelerate business processes by 30% to 50% at scale

• Low-value work drops 25%-40% - The same deployments show agents reduce time on low-value tasks by 25% to 40%, freeing capacity for higher-judgment work

• Agentic Business Transformation coined - Ema and HFS Research introduced ABT as a formal category superseding digital transformation and co-pilot initiatives

The Bigger Picture: The 76% problem is not a technology gap. It is a management and structural one. Companies that drop AI agents into org charts designed for humans will see the same failure modes as every digital transformation wave: pilots that work in isolation, then stall at scale. Organizations that get ahead will redesign workflows, decision rights, and success metrics for a world where agents are active participants. Companies spending on AI now should be asking not just what agents can do, but whether their organization is built for what agents need.

Read More →

🤖 AI Agents Force a Web Rebuild

Bots now account for 31% of all web traffic per Cloudflare, with AI crawlers making up a quarter of that volume. AWS launched serverless search infrastructure built for AI agent workloads, which spike without warning and scale to zero when idle. Non-human traffic will surpass human traffic by mid-2027.

Read More →

🔍 Claude 4.8 Catches Its Own Mistakes

Anthropic released Claude Opus 4.8 with improved honesty during complex tasks. The model is 4 times less likely than its predecessor to let flaws in its own code pass without flagging them. Users can also adjust how much effort Claude applies per task, balancing speed against thoroughness.

Read More →

🎯 OpenAI Tax Agent Hits 97% Accuracy

OpenAI's Codex helped build a self-improving tax agent that processed 7,000 returns with up to 97% accuracy. At launch, only 25% of returns reached 75% correct completion; within six weeks, 86% hit that mark. The system saves practitioners about a third of their time without any engineer intervention.

Read More →

Microsoft 365 Copilot ⚡

Microsoft just relaunched 365 Copilot with a cleaner interface that loads twice as fast. A progressive disclosure feature surfaces tools and controls based on your specific prompt, reducing clutter. The prompt box expands as you type and supports text formatting directly inside it.

Try it →

Warp 🖥️

Warp is an AI-native terminal where agents now co-create around 90% of the company's own pull requests. Open-sourced this year with GPT-5.5 as the default model for long-horizon coding tasks - GPT-5.5 uses 30% fewer tokens per agentic task than the prior version. Used by more than 56% of Fortune 500 companies.

Try it →