Date: {{current_date_full_with_day}}

Hey {{first_name | AI enthusiast}},

AI is no longer just answering questions. It is making decisions, moving capital, and shaping taste in real time.

This edition captures a shift that feels structural, not incremental.

• First, AI agents are gaining wallets. Coinbase is introducing agentic wallets, giving software the ability to hold and spend digital assets onchain.

• At the same time, capital is flooding the frontier. Anthropic is raising $30 billion, pushing its valuation to $380 billion.

📈 That scale of funding signals long-term conviction. Investors are betting that foundation models are becoming core infrastructure.

• On the model side, Google is advancing reasoning with Gemini 3 Deep Think. The focus is shifting from speed to structured thinking.

🧠 Meanwhile, MiniMax is launching M2.5 and M2.5 Lightning. Open models are moving closer to frontier performance.

Then comes a quieter reminder from Paul Graham. In a world of stronger tools, taste is still separating average work from lasting work.

Three forces are converging: autonomous agents, massive capital, and improving open models.

Hope you enjoy this power-packed edition!

— Renjit Philip

PS: If you want to unleash the power of AI agents to grow your business, setup time speak to me, here»

• The AI funding race just hit a new level. Anthropic raised $30 billion in a Series G round, reaching a $380 billion post-money valuation.

📈 This round marked one of the largest private fundraises in tech. It signaled strong investor conviction in frontier AI models and long-term platform plays.

• The company said it will use the capital to scale its AI systems. That includes advancing its Claude models and expanding compute capacity.

🧠 Leadership framed the raise as fuel for research and safety. The focus stayed on building powerful systems with clear guardrails.

• The valuation placed Anthropic among the most valuable private companies in the world. It showed how fast AI firms can compound value.

⚙️ The funding also supported infrastructure growth. Training advanced models demands massive compute and specialized hardware.

• The company positioned this round as a step toward long-term resilience. It reinforced its ambition to compete at the highest level of AI development.

🌍 The broader signal was hard to miss. Capital continues to concentrate around a few leading AI labs. For founders and operators, the message is clear. Platform AI is not slowing down; it is scaling fast.

🚀 When capital moves at this scale, strategy matters more than ever. Remember how the SaaS companies stocks collapsed when Anthropic announced the launch of Claude Co-work? The valuation justifies the launch.

• What happens when an AI model pauses to think before answering? That question shaped the release of Gemini 3 Deep Think from Google.

🧠 The new system builds on the Gemini family. It focuses on stronger reasoning, especially for complex tasks that require step-by-step thinking.

• Instead of rushing to a response, the model takes extra compute time. It evaluates different possible answers before selecting the best one.

⚙️ Google described this as using more inference-time compute. In simple terms, the model thinks longer to improve quality.

• This approach helps in coding, math, and multi-step problem solving. The company shared examples where deeper reasoning improved accuracy.

📊 The update reflects a broader shift in AI research. Labs now compete not just on size, but on how well models reason.

• Gemini 3 Deep Think also shows how companies balance speed and depth. Faster answers work for chat; deeper thinking matters for harder tasks.

🔍 Google positioned this as part of its long-term AI roadmap. The goal is to build systems that reason more like humans.

• Performance gains now come from smarter inference strategies, not only bigger models. The race is about how well AI systems can think before they speak.

• Why do some founders build products people love, while others build products people ignore? The answer, Paul Graham argues, often comes down to taste.

🎯 Taste is not about style or decoration. It is about knowing what is good, and caring enough to reject what is not.

• Graham explains that taste means noticing small flaws. A clunky sentence, a slow feature, a messy layout; these details matter more than people admit.

🧠 He points out that many creators lack taste early on. They cannot see what is wrong with their own work.

• The gap between taste and skill creates frustration. You know something feels off, but you cannot yet fix it.

⚙️ The way forward is simple but hard. Do a lot of work. Study examples of great work. Compare, refine, repeat.

• Graham stresses that taste improves through exposure. You train it by reading good writing, using strong products, and studying clean design.

📚 He also notes that high standards act like a compass. They guide decisions when there is no clear rulebook.

• In startups, taste shows up everywhere. It shapes product decisions, hiring choices, even how you write your landing page.

🚀 The core message is practical. If you want to build something great, raise your standards and sharpen your eye. Taste is not loud. It does not announce itself. But over time, it separates average work from work that endures.

• What if your wallet could do more than store assets? Coinbase introduced agentic wallets to push that idea forward.

🤖 These wallets are designed for AI agents, not just humans. They allow software agents to hold assets and transact onchain.

• The goal is simple. Give AI systems the ability to pay, receive funds, and interact with decentralized applications.

🔐 Coinbase built this within its developer platform. It focused on programmable, secure infrastructure for agent-driven transactions.

• Instead of manual clicks, an AI agent can execute payments as part of a workflow. For example, it could pay for services or access APIs automatically.

⚙️ The wallet integrates with onchain identity and smart contracts. That means agents can verify actions and follow predefined rules.

• This moves crypto from passive storage to active participation. The wallet becomes part of an automated system.

🌐 For developers, this opens new design patterns. They can build apps where agents manage subscriptions, settle payments, or coordinate tasks.

• The broader shift is clear. As AI agents grow more capable, they need financial rails that match their autonomy. Remember how OpenClaw (we wrote about it recently), creates agents that work 24×7 and in some cases were connected to bank accounts to pay for stuff?

🚀 Agentic wallets point to a future where software does not just think; it transacts.

• The open model race is moving fast. MiniMax introduced M2.5 as a new generation model built to compete at the top tier.

🧠 The company designed M2.5 to improve reasoning and overall capability. It aimed to close the gap with leading frontier systems.

• MiniMax highlighted performance gains across key benchmarks. The model showed stronger results in complex tasks compared to earlier versions.

⚙️ The release also focused on efficiency. The team worked to balance compute demands with practical deployment needs.

• M2.5 builds on prior iterations, refining architecture and training methods. This reflects a steady improvement cycle rather than a one-off leap.

📊 The announcement positioned M2.5 as part of a long-term roadmap. It signals intent to remain competitive in a crowded AI landscape.

• For developers, the appeal is clear. Stronger open models mean more flexibility and lower dependence on closed platforms.

🚀 The broader message is simple. Open AI systems are gaining ground, and the performance gap continues to narrow.