Get tomorrow's brief in your inbox
Today: OpenAI landed $110 billion from Amazon, Nvidia, and SoftBank, hitting 900 million weekly users. The Pentagon banned Anthropic after the company refused to power autonomous weapons. AI coding agents now port entire Python libraries to Rust in days, not months.
OpenAI closed a $110 billion funding round that reshapes the AI infrastructure landscape. Amazon is investing up to $50 billion (starting with $15 billion, the rest tied to milestones including a potential IPO), while Nvidia and SoftBank each committed $30 billion. The company now claims 900 million weekly active users and 50 million paid subscribers, up from 800 million weekly users in October 2025.
For small businesses, this matters because the Amazon deal extends beyond capital. The two companies are building custom models for Amazon's consumer products (including a major Siri-like Alexa upgrade), and OpenAI's new enterprise platform called Frontier will run on AWS using Amazon's Trainium chips. OpenAI will spend $100 billion on AWS over eight years. This means tighter integration between the tools you already use (AWS infrastructure, Alexa) and OpenAI's models. If your business runs on AWS, expect more AI capabilities to arrive as built-in options rather than third-party integrations.
The funding also signals OpenAI's shift away from Microsoft exclusivity. While both companies insist their partnership remains "strong and central," OpenAI is diversifying. Microsoft still holds its exclusive license to OpenAI's intellectual property, but the Amazon deal shows OpenAI is no longer locked into a single cloud provider. For businesses evaluating AI platforms, this means more deployment options and potentially better pricing as cloud providers compete. OpenAI, TechCrunch, The Verge, CNBC
Perplexity Computer - Perplexity's new agentic tool unifies 19 different AI models to handle complex workflows like collecting data, creating analysis, and generating finished websites or visualizations. Available only on the $200/month Perplexity Max subscription tier. The tool runs entirely in the cloud and can create subagents to handle specific problems. Perplexity is pivoting from mass-market search to "boutique" enterprise research for users making "GDP-moving decisions." If you're doing deep research work that involves pulling data from multiple sources and synthesizing analysis, this might justify the premium tier. But at $200/month, you need to be doing a lot of research to make it worthwhile. TechCrunch
Amazon Bedrock Stateful Runtime - Amazon and OpenAI launched a stateful runtime environment for AI agents in Amazon Bedrock. This brings persistent memory and orchestration to multi-step workflows, meaning your agents can remember context across sessions and handle longer, more complex tasks. If you're building automation that requires multiple steps over time (like order processing that spans days, or customer support that tracks issues across interactions), this infrastructure upgrade makes those workflows more reliable. OpenAI
Free Claude Max for open source maintainers - Anthropic is offering free $200/month Claude Max subscriptions to open source maintainers of projects with 5,000+ GitHub stars or 1 million+ monthly NPM downloads. The offer lasts six months and is capped at 10,000 contributors. If you maintain a widely-used open source project, this is essentially $1,200 in free AI access. Apply even if you don't quite meet the criteria; Anthropic says they'll consider projects "the ecosystem quietly depends on." Simon Willison
Google Gemini Deep Think upgrade - Google upgraded Gemini 3 Deep Think with enhanced reasoning for complex science, research, and engineering problems. It can process messy, incomplete data and extract actionable insights. Available to Google AI Ultra subscribers through the Gemini app, with API access rolling out to researchers and enterprises. If you work with inconsistent data sources (production reports, consumer feedback from multiple channels), Deep Think can help you analyze and optimize operations even when the data isn't perfect. The subscription cost may be a barrier for small teams, so evaluate whether the time savings justify the expense. Small Business Trends
AI saves 2.6 hours per day - A GoTo study found that small business owners lose 2.6 hours daily to tasks AI could streamline. That adds up to 13 hours per week. Content creators interviewed by Lenovo report using ChatGPT to analyze video performance data (condensing hours of investigation into minutes), schedule tasks, simplify meal prep, and speed up design workflows for logos and social media. One MBA student said, "I use it in every single portion of my life." The pattern: use AI to offload repetitive tasks, then redirect that time to strategic planning, customer engagement, or creative work. Start with one high-friction task (data analysis, scheduling, content drafting) and measure the time saved before expanding to other areas. Small Business Trends
AI coding agents are production-ready - Developer Max Woolf documented his shift from AI coding skeptic to believer after using agents to port Python's scikit-learn library to Rust, complete with performance optimizations that beat the original implementations. His conclusion: "Opus 4.6 and Codex 5.3 are an order of magnitude better than coding LLMs released just months before." The frustration, he notes, is that "it's impossible to publicly say this without sounding like an AI hype booster, but it's the counterintuitive truth." For small businesses, this means complex software projects that would have required hiring specialized developers are now feasible with AI assistance. If you've been putting off custom tools or integrations because of cost, reevaluate. Simon Willison
Tariffs and AI-powered pricing models - Dirk Buikema, founder of Hyperwear (an Inc. 5000 company), used ChatGPT to model the impact of tariff-driven price hikes before implementing them. Generative AI let him run multiple scenarios quickly, understanding customer response and margin impact before committing to new pricing. If you're facing cost pressures (tariffs, supply chain issues, inflation), use AI to model different pricing strategies and forecast customer behavior. It's faster and cheaper than hiring a consultant, and you can iterate in real time. Inc.com
Zapier vs. n8n cost comparison - Zapier published a detailed comparison positioning itself against n8n, the self-hosted automation platform. The core difference: Zapier is fully managed cloud infrastructure with transparent pricing, while n8n requires server provisioning, security implementation, and JavaScript/Python expertise. Zapier's argument is that "self-hosted" doesn't actually keep everything on-premises since most integrations require external API calls anyway. For small businesses, the choice is clear: if you have technical teams who want infrastructure control and have time to maintain it, n8n might save money. If you want to ship automation quickly without operational overhead, Zapier's higher upfront cost avoids hidden infrastructure and maintenance expenses. Zapier
Suno hits $300M in annual recurring revenue - AI music generator Suno now has 2 million paid subscribers and $300 million in annual recurring revenue, up from $200 million just three months ago. Suno lets users create music from natural language prompts. Warner Music Group settled its copyright lawsuit and reached a licensing deal, signaling that AI-generated music is moving from legal gray area to legitimate business tool. If you create video content, podcasts, or ads and need background music, Suno offers a low-cost alternative to stock music libraries. Just be aware that AI-generated music can sound generic, and some platforms or clients may require human-created soundtracks. TechCrunch
Pentagon bans Anthropic, OpenAI steps in - President Trump directed federal agencies to stop using Anthropic products after the company refused to allow its models to power mass domestic surveillance or fully autonomous weapons. The Pentagon designated Anthropic a supply-chain risk, meaning no military contractor can do business with the company. OpenAI moved quickly to fill the gap, announcing a Pentagon deal that preserves the same restrictions Anthropic fought for. For businesses selling to government or defense contractors, this is a reminder that AI vendor choices now carry political and compliance implications. If your client base includes federal contractors, verify that your AI tools are approved for their use. TechCrunch
Apple sits out the AI spending race - Apple spent $12.72 billion on AI-related capex in fiscal 2025, far less than Alphabet, Meta, Microsoft, and Amazon (which spent $30 billion to $125 billion in the same period). Analysts aren't worried. Apple's advantages include a tightly integrated operating system, a massive installed base, and a strong privacy reputation. The company inked a deal with Google to power its AI features, including a major Siri upgrade expected later this year. For small businesses, this means Apple devices will get smarter without Apple building its own foundation models. If your team uses iPhones and Macs, expect better AI integration across devices this year without switching ecosystems. CNBC
Samsung's AI camera crosses the line - Samsung's Galaxy S26 introduces AI camera features that automatically alter photos in real time, raising questions about what constitutes a "photo" anymore. The Verge called it a "photography nightmare" and "a crossing of a line" in the ongoing "what is a photo?" debate. For businesses that rely on authentic visual content (real estate, product photography, documentation), this trend toward AI-manipulated imagery creates trust problems. If you're using AI to enhance photos, be transparent about it. Customers and clients are growing more skeptical of visual content that looks too polished or perfect. The Verge
Binary search over HTTP range requests - Developer Simon Willison built a Unicode Explorer tool that uses HTTP range requests and binary search to query a 76.6MB file without downloading the whole thing. The technique works for any large, sorted dataset where you need to look up specific records quickly. If you're building tools that query large reference files (product catalogs, databases, documentation), this approach can drastically reduce bandwidth and load times. Willison documented the entire build process, including the Claude Code prompts he used. Simon Willison
Don't use passkeys to encrypt user data - Security expert Tim Cappalli is urging the identity industry to stop using passkeys for data encryption. Users lose their passkeys frequently and may not understand that their data has been irreversibly encrypted and can no longer be recovered. Passkeys are excellent for authentication, but terrible for encryption. If you're building apps that store user data, use passkeys for login security, but encrypt data with recoverable keys or offer clear backup options. Simon Willison
The infrastructure deals this week (OpenAI-Amazon, Bedrock stateful runtime, Gemini Deep Think) all point to the same trend: AI is moving from experimental tools to embedded infrastructure. Businesses that wait for "AI to mature" are missing the point. It's already here, already integrated into the platforms you use, and already being used by competitors to save time and reduce costs. The question isn't whether to use AI, but which workflows to automate first and how to measure the return.
Watch for more enterprise AI announcements from Google and Microsoft as they respond to the OpenAI-Amazon deal. Also watch for fallout from the Pentagon-Anthropic dispute. If other agencies or contractors follow the supply-chain risk designation, it could fragment the AI vendor landscape along political lines.