Enterprise AI loyalty is a myth, and the web is already half-AI
Businesses are bouncing between OpenAI and Anthropic like ping-pong balls, a third of the web is now AI-written, and ChatGPT wants to text your friends.
Subscribe free →Enterprise AI Has Zero Brand Loyalty — and That's a Problem for Everyone
New data shows OpenAI is clawing back business users it lost to Anthropic, but the more alarming takeaway is how fast companies are switching. Businesses are hopping between AI providers every time a new model drops, which means neither OpenAI nor Anthropic has built anything resembling a moat. If your entire competitive advantage can be erased by a rival's next model release, "enterprise AI" is basically a commodity business with a VC-scale burn rate.
Why it matters: Sticky enterprise software typically has switching costs baked in — AI APIs apparently don't, which is going to make the economics of this industry brutal over the long run.
A Third of the Web Written Since ChatGPT Launched Is AI-Generated
A new study found that roughly one-third of web pages published since ChatGPT's November 2022 launch show signs of AI authorship. That's not a fringe phenomenon — it's a wholesale transformation of what the internet actually is. The implications cascade everywhere: AI training data is increasingly AI-generated, search quality is degrading, and the feedback loop between models and the content they produce is tightening fast.
Why it matters: If AI models are being trained on AI-written content at scale, we're already deep into a data laundering spiral that nobody fully understands how to measure or fix.
ChatGPT Will Now Text Your Friends for You
OpenAI has launched an Apple Messages integration that lets ChatGPT compose and send iMessages on your behalf. You tell it what you want to say, it drafts and sends — essentially turning the AI into a personal communications proxy. This is one of the more mundane-sounding features to cross a genuinely scary threshold: an AI agent with unsupervised write access to your social relationships.
Why it matters: Agentic AI moving from "answer my question" to "act on my behalf in the real world" is the pivotal shift in this technology's threat model — and it's happening through a texting feature, not a dramatic announcement.
Google Throws Publishers a Lifeline as AI Search Kills Their Traffic
Google is rolling out a "follow" button that lets readers designate preferred news sources, which then get prioritized across Search, Discover, and Google News. Translation: Google helped break publisher traffic with AI Overviews and is now offering a preference dial as a remedy. It's a genuinely useful feature, but let's be clear — Google is handing publishers a band-aid after performing the surgery.
Why it matters: This sets a precedent for how platforms may try to manage the political fallout of AI-driven traffic destruction without actually reversing it.
AI Training Data Startup Micro1 Hits $500M Revenue Run Rate
Micro1, which supplies human-generated and human-verified data for AI training, has hit a $500 million gross run rate — a number that would have seemed absurd for a data labeling company just three years ago. The AI training boom is minting a whole new category of infrastructure businesses, and demand is outpacing supply fast enough that multiple rivals are reportedly on similar trajectories.
Why it matters: The economics of AI training data are starting to look like the early cloud wars — whoever locks in the best data pipelines now may have a durable advantage when the market consolidates.
YouTube's Top Creators Are Getting Torched for Promoting AI Video Tools
Prominent filmmaking YouTubers Matti Haapoja and Sam "Kold" Kolder posted sponsored content showcasing Higgsfield's Seedance 2.5 AI video generation platform — and the backlash from the creator community was immediate and fierce. Other creators called it a betrayal, arguing that promoting tools that automate video production undermines the very craft that built their audiences. The tension between creator economics and creator solidarity is now fully in the open.
Why it matters: When the people who built their careers teaching filmmaking start getting paid to promote AI that replaces filmmakers, the cultural reckoning around AI and creative labor gets a very public face.
Quick Hits
- →Grok started sending users gibberish responses this week — affected users were on Grok Lite and noticed issues Wednesday morning. TechCrunch
- →Waymo doubled its lobbying spend as it fights Uber for regulatory clearance to scale fully autonomous robotaxis nationwide. Ars Technica
- →Tesla, Uber, and Waymo all got permits to deploy up to 8,000 robotaxis combined in Nevada over the next 12 months. TechCrunch
- →Starcloud raised $250M to build orbital data centers as ground-based launch access gets more competitive. TechCrunch
- →AI coding tools are quietly reshaping language popularity — TypeScript became GitHub's most-used language in August 2025, the biggest ranking shift in a decade. AI News
AI TLDR