Big Creek Briefing

AI Companies Blew Their Cover

Welcome to this week's Big Creek Briefing.

This week, the AI industry made headlines by accident, and the story behind those headlines is worth paying close attention to if you run a business. Anthropic leaked its own source code, confirmed the existence of a next-generation autonomous AI model it wasn't ready to announce, and Perplexity got hit with a class-action lawsuit for secretly sharing your conversations with Meta and Google. Meanwhile, OpenAI closed a record $122 billion funding round at an $852 billion valuation, signaling that the money behind this technology is still accelerating. For founders and small business owners, the lesson this week is not just about the tools themselves. It is about trust, data, and what it means to be watching closely.

Let's jump in!

Big Stories of the Week

OpenAI Closes a Record $122 Billion Round at an $852 Billion Valuation: OpenAI just closed the largest private funding round in history, pulling in $122 billion at a valuation of $852 billion, with a likely IPO window opening in Q4 2026. The company is now generating $2 billion in revenue per month, crossing $25 billion in annualized revenue after growing from just $6 billion at the end of 2024. For founders, this signals something practical: the AI infrastructure you are building on is no longer a speculative bet, it is becoming one of the most well-capitalized industries in history.
Anthropic Leaks Claude Code Source Code and Accidentally Confirms Claude Mythos: In a significant slip, Anthropic mistakenly published 1,906 internal files detailing Claude Code's tool orchestration logic, internal APIs, and permission systems to a public developer registry. Separately, an accidental data leak forced Anthropic to confirm the existence of Claude Mythos, an unreleased model that plans and executes sequences of actions autonomously across systems without waiting for human input at each step. No customer data was exposed, but the incident raises real questions about operational security at even the most safety-focused labs.
Perplexity AI Faces Class-Action Lawsuit Over Secret Data Sharing: A proposed class-action lawsuit filed this week accuses Perplexity AI of using undetectable tracking technology to share user conversations with Meta and Google, potentially in violation of California privacy laws. According to the complaint, trackers are downloaded to users' devices the moment they log into Perplexity's homepage, giving Meta and Google access to what users type into the AI search engine. For any business owner using AI tools that handle customer data or sensitive internal conversations, this is a reminder to read the fine print before routing real business information through consumer AI products.

Main Insights

AI in SMB's: By the Numbers

The SMB AI adoption numbers from early 2026 tell a clear story: 42% of small and mid-sized businesses now use AI in at least one core business process, up sharply from just 23% in 2024. But the more nuanced finding is that 74% of SMBs are already using AI indirectly through embedded features in tools they already pay for, such as email filtering, CRM lead scoring, and automated reporting. The average SMB is now spending $18,000 per year on AI-related tools and subscriptions, a number that is rising fast. The practical takeaway for founders: you are likely already paying for AI capabilities inside your existing software stack that you have never turned on. Before purchasing a new AI tool this quarter, spend one hour auditing what the tools you already use can actually do.

Founder Insights

The most important shift happening right now for founders is not which AI model is most powerful. It is the move from generic tools to niche, vertical AI solutions built for specific industries and workflows. Multiple reports this quarter point to the same pattern: small businesses that are winning with AI are not using broad tools for everything. They are identifying one or two high-friction processes in their specific business and adopting AI built specifically for that problem. The verticals showing the fastest SMB adoption include contract generation, AI-assisted hiring, dynamic pricing, and customer churn prediction. For founders building products, this is also a market signal worth taking seriously. The foundational model race is largely settled for now. The next wave of value, and the next wave of fundable companies, will be built on top of those models in focused, industry-specific layers.

Quick Hit: One Prompt to Simplify Your Week

"I currently subscribe to the following business tools: [list your tools, e.g. HubSpot, Google Workspace, Slack, QuickBooks, etc.]. For each tool, tell me: (1) what AI features are currently included in my plan that I may not be using, (2) what tasks those features can automate or accelerate, and (3) one specific thing I should try this week with each. Be specific and practical, not general."

AI Tool Spotlight: Taskade

If you have ever wished your task manager, project board, document workspace, and AI assistant all lived in the same place and actually talked to each other, Taskade is worth a serious look. It is a collaborative productivity platform built around AI agents, combining kanban boards, mind maps, documents, and automated workflows in a single app. What makes it different from a standard project management tool is the AI agent layer: you can build custom AI agents that understand your specific projects, run research tasks, generate content, and trigger automations, all without writing a single line of code. For founders who are juggling multiple initiatives and need their AI to actually know what they are working on, the context-aware agents are a genuine step up from copy-pasting prompts into a separate chat window.

The practical use case for small business owners is around project kickoff and planning. Rather than starting a new project from a blank screen, you describe your goal to a Taskade agent and it generates a structured task breakdown, suggests relevant templates, and can even kick off research on competitors or market context in parallel. Teams use it to run async standups, manage client deliverables, and build internal knowledge bases that the AI can query on demand. The "prompt-to-app" builder, part of its Genesis feature set, lets you deploy lightweight internal tools without a developer, which is a real advantage for lean founding teams.

Pricing starts at free with a capable feature set for solo users. The Starter plan is $8 per month and covers most solo founder needs with 1,000 AI generations monthly. The Plus plan at $16 per month unlocks the full AI agent builder and automation triggers, and the Pro plan at $39 per month supports multi-agent collaboration and larger teams. For a 10-person team, Taskade Pro comes out to roughly $20 per month total, which undercuts most comparable tools significantly. You can find it at taskade.com.

What We're Watching

GPT-5.4 and the 1 Million Token Context Window

OpenAI's API now supports a 1 million token context window through GPT-5.4, meaning developers can feed an entire business's documents, contracts, or customer history into a single AI request. This will start enabling use cases in the next few months that weren't possible before.

The AI Data Privacy Reckoning

Google's new Gemini 3.1 Flash-Lite is priced at just $0.25 per million input tokens, 2.5x faster than its predecessor. The cost of building and running AI-powered products is dropping fast.

Samsung Targets 800 Million Gemini-Powered Devices

Samsung announced a goal to double its Gemini AI footprint to 800 million devices by end of 2026, spanning mid-tier and budget smartphones globally. When AI assistants are standard on 800 million phones, consumer behavior around search, shopping, and business discovery will start shifting fast.

Federal Judge Blocks Government AI Ban

A U.S. federal judge ruled this week that the Trump administration violated free-speech protections by attempting to ban the use of Anthropic's AI models in government systems. The ruling signals that government AI policy is still very much in flux and could affect procurement, compliance, and partnership opportunities for businesses working with public sector clients.

The Wrap(TLDR)

This week was a case study in how fast things move in AI when no one is looking. Anthropic leaked its own source code and confirmed a new autonomous model it wasn't ready to announce. Perplexity got sued for secretly sharing your conversations. And OpenAI raised $122 billion, solidifying its place as the most well-capitalized private company in history. The data signal for founders is practical: SMB AI adoption has nearly doubled in two years, but most businesses are still leaving paid AI features unused. Start there before buying anything new.

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.