Big Creek Briefing

The AI Arms Race and Small Business

Welcome to this week's Big Creek Briefing.

This week in AI felt less like incremental updates and more like tectonic shifts. Google dropped a new open-source model family with agentic capabilities running directly on your phone, Anthropic confirmed it's hiding its most powerful model from the public, and a blockbuster PwC study revealed that 74% of AI's economic value is flowing to just 20% of companies. For founders and small business owners, the message is clear: the window to get ahead of the curve is narrowing, and the tools to do it have never been more accessible or more powerful. Here's what you need to know this week.

Let's jump in!

Big Stories of the Week

Google Releases New AI Agent to Challenge OpenAI and Anthropic: On April 22, Google released Gemma 4, its strongest open-weight model family yet, under an Apache 2.0 license, meaning anyone can use it freely. The models range from 2B to 31B parameters and come with native agentic capabilities: multi-step planning, autonomous action, function-calling, and offline operation on Android devices. For founders who want to build automation into their products or workflows without paying per-API-call, this is a major unlock.
Anthropic's Most Powerful Model Will Not Be Public:
Anthropic confirmed the existence of Claude Mythos, described internally as a "step change in capabilities" representing the most advanced model it has ever built. Rather than releasing it broadly, Anthropic is gating access to only 50 organizations through a program called Project Glasswing. This signals a new era where frontier AI capability may increasingly sit behind closed doors, and the companies with relationships or funding to access these private programs will have a real competitive edge.
OpenAI, Anthropic, and Google Formally Unite Against Chinese AI Theft: The three biggest names in AI announced they are now sharing intelligence through the Frontier Model Forum to stop Chinese AI companies from copying their models through adversarial distillation. This is the first formal alignment between these otherwise fiercely competing labs on a security issue. For founders building on top of these platforms, this signals that IP and model integrity are becoming a serious industry concern, with downstream implications for how APIs and licensing terms may evolve.

Main Insights

How Many Companies are Really Using AI?

A new PwC global study of 1,217 senior executives found that 74% of AI's economic value is being captured by just 20% of companies, and those top performers are generating 7.2 times more AI-driven revenue and efficiency gains than the average competitor. The key differentiator is not using more AI tools but using AI to reinvent the business model and pursue growth from industry convergence, not just to cut costs. For small business owners, this is a wake-up call: efficiency gains alone will not put you in the winner's tier. The founders who are winning are using AI to open new revenue streams, not just to move faster on old ones. The gap between AI leaders and laggards is widening fast, and the time to move from "experimenting with AI" to "building your business around AI" is right now.

Founder Insights

The AI funding environment in 2026 is the best it has ever been for early-stage founders in absolute terms, with seed and Series A AI rounds commanding higher valuations than any previous period. However, a clear pattern is emerging in where capital is concentrating: vertical AI wins. Horizontal, general-purpose tools are oversaturated, while startups doing one specific thing exceptionally well for a defined audience are pulling ahead. Voice AI agents replacing call centers, legal AI cutting document review time by 80%, and healthcare AI triaging patient intake are all commanding premium valuations and real revenue. If you are building or investing in AI right now, the most important strategic question is not "what can AI do?" but "what single problem can I solve better than anyone else for a specific group of people?"

Quick Hit: One Prompt to Simplify Your Week

"I run a [type of business] serving [target customer]. My three biggest competitors are [A, B, C]. Based on how AI is transforming this industry in 2026, identify: (1) where AI is most likely to commoditize what I currently charge for, (2) one new revenue stream I could open using AI that my competitors are unlikely to pursue in the next 12 months, and (3) what I should stop doing manually that I can fully automate within 30 days."

This prompt forces your AI assistant to think strategically about your specific business instead of giving generic advice. Run it monthly and compare the outputs.

AI Tool Spotlight: n8n

If you have ever felt like Zapier costs too much once your automations scale, or you wanted more control over what your workflows actually do, n8n is the tool that technical founders and operations-savvy business owners are quietly building their entire stack around in 2026.

n8n (pronounced "nodemation") is an open-source workflow automation platform with a visual drag-and-drop canvas where you connect apps, triggers, data transformations, and logic operations to build automated workflows. What sets it apart in 2026 is its native AI layer: the platform ships with 70+ LangChain-based nodes for building AI agents, connecting vector databases, routing LLM calls, and managing conversation memory. That means you can build a customer intake agent that reads incoming emails, classifies leads, updates your CRM, sends a personalized reply, and logs the whole thing to a Google Sheet, without writing a single line of code. It is genuinely the kind of automation that would have cost a full-time developer salary three years ago.

Pricing is where n8n stands out from competitors like Zapier or Make. The self-hosted version is free forever for most use cases, and founders running more than 5,000 workflow executions per month report saving three to twenty times what they were spending on Zapier at equivalent scale. The cloud-hosted version starts at around $20/month for small teams that prefer not to manage their own server. Fair warning: n8n has a steeper learning curve than Zapier for non-technical users, but if you can follow a flowchart, you can build powerful automations. Find it at n8n.io.

What We're Watching

On-device AI going mainstream

With Gemma 4 running agentic workflows natively on Android, the era of AI that works offline and keeps your data local is arriving faster than expected, with major implications for privacy-conscious industries.

The private model tier

Claude Mythos joining OpenAI's operator-tier models behind paywalls and access programs suggests a two-class system is emerging between publicly available AI and frontier-capability AI reserved for select partners.

The 80/20 AI business gap widening

The PwC findings are not a one-year anomaly. As AI leaders compound their advantages, small businesses that stay in "exploration mode" risk finding the gap impossible to close within 18 to 24 months.

Agentic AI hitting mobile

Google's Agent Skills running autonomously on Android phones signals that AI agents will increasingly live in your pocket, not just in browser tabs, reshaping how and when business owners interact with their tools.

The Wrap(TLDR)

Google went open-source with a powerful new agentic AI model family this week, Anthropic is keeping its best model locked away from the public, and a major PwC study confirmed that 74% of AI's economic gains are flowing to just 20% of companies. The pattern is clear: the founders winning with AI are not just using more tools, they are using AI to reinvent how their business makes money. If you are still in experimentation mode, it is time to pick a lane and build.

If you want help designing AI workflows or understanding how these new models fit into your business, Big Creek Growth can help. Reply to this email and we will walk through it together.