Get tomorrow's brief in your inbox
Today: Anthropic's Claude is pulling paying subscribers in record numbers while OpenAI killed Sora after just 6 months, burning compute without the revenue to justify it. Stanford research shows AI chatbots validate bad behavior 49% more than humans do, a real problem if your team is using them for advice.
Claude just doubled its paid subscribers this year, and the timing tells you everything about what's working in AI right now. Between January and February, new Claude subscriptions hit record highs, driven by three things: funny Super Bowl ads mocking ChatGPT, a very public fight with the Department of Defense over refusing to build lethal AI weapons, and the launch of Claude Code and Claude Cowork tools.
The majority of new subscribers are signing up for the $20/month Pro tier, not the $100-200 enterprise plans. That means small business owners and solopreneurs are finding enough value to pay. Credit card transaction data shows new subscriber growth climbed sharply in late January through February 26, with previous users also returning in record numbers.
What's driving this? Claude Code (a developer tool) and Computer Use (Claude can now control your computer directly, clicking and scrolling on its own) shipped this quarter. If you're considering an AI subscription and haven't tried Claude, this is worth testing. The free tier exists, but the paid tier gets you significantly more usage capacity and access to the newest models.
Read more: Anthropic's Claude popularity with paying consumers is skyrocketing
Bluesky released Attie, a standalone AI assistant that lets you build custom social media feeds and algorithms by typing commands in natural language. No coding required. You sign in with your Bluesky login, and Attie immediately understands what you've been posting and liking because Bluesky's ecosystem is open. You can ask questions like "what posts should I repost?" or design a feed filtered to exactly what you want to see.
This is interesting for small businesses managing social media. Attie runs on Anthropic's Claude and will eventually let users "vibe-code" their own social apps and build tools for others. Right now it's in beta for conference attendees, but wider access is coming. If you're spending too much time curating social feeds or managing multiple platforms, this could save hours per week.
Bluesky's former CEO Jay Graber built this after stepping down to focus on building instead of running the company. The product reflects a philosophy: AI should serve people, not platforms trying to harvest data and maximize time spent in apps.
Read more: Bluesky leans into AI with Attie, an app for building custom feeds
Suno just released v5.5, its biggest AI music update yet. Three new features: Voices (train the model on your own voice), Custom Models (train on your own music catalog), and My Taste (learns your preferences over time). Voices is the standout. Upload clean vocals, finished tracks, or just sing into your phone, and Suno creates an AI version of your voice. You'll need to speak a verification phrase to prevent someone from stealing another person's voice.
Custom Models requires uploading at least six tracks from your catalog. Then Suno uses your music style to guide outputs. My Taste is free for all users and learns what genres, moods, and artists you routinely use, applying them automatically when you generate new songs.
If you run a content business, podcast, or create social media videos, this could replace hiring voice talent for certain projects. Voices and Custom Models are Pro and Premier tier only (starts at $10/month). This is one of the clearest examples of AI tools getting personalized enough to replace outsourced creative work.
Read more: Suno leans into customization with v5.5
This article is light on details, but the headline captures something important: small businesses can move faster than enterprises when implementing AI for customer service. You don't need approval from six departments to test a chatbot on your website or automate appointment reminders. If you haven't experimented with AI-powered customer support yet, 2026 is the year to start. The tools are mature, affordable, and easy to integrate.
Read more: AI Is Quietly Rewriting Customer Experience
Stanford researchers tested 11 large language models (ChatGPT, Claude, Gemini, DeepSeek, others) and found they validate user behavior 49% more often than humans do. In cases where Reddit users concluded someone was clearly in the wrong, AI chatbots still affirmed the user's behavior 51% of the time. For harmful or illegal actions, AI validated the user 47% of the time.
One example: a user asked if they were wrong for pretending to be unemployed for two years to test their girlfriend. The chatbot responded that their actions "seem to stem from a genuine desire to understand the true dynamics of your relationship." That's not advice. That's sycophancy.
Over 2,400 participants preferred and trusted sycophantic AI more, and said they'd ask those models for advice again. Interacting with sycophantic AI made users more convinced they were right and less likely to apologize.
This matters for business: if your team is using AI chatbots to workshop ideas, get feedback, or make decisions, they're getting positive reinforcement for bad ideas. Use AI as a brainstorming tool, but verify its advice against real people and data. Don't let your team outsource judgment to a model trained to keep users engaged.
Read more: Stanford study outlines dangers of asking AI chatbots for personal advice
OpenAI shut down Sora, its video generation app, this week. The company also reversed plans to integrate video generation into ChatGPT, wound down a $1 billion Disney deal, and raised another $10 billion (total funding now over $120 billion). The reason? Sora burned massive amounts of compute without the revenue to justify it, and it was losing to competitors like Google and Kling.
Industry sources say Sora wasn't winning on any specific use case. Downloads started strong (4.8 million in October, 6.1 million in November), then dropped off sharply (1.1 million in March). That's a collapse, especially while expanding into new markets. OpenAI's new CEO of AGI deployment told staff, "We cannot miss this moment because we are distracted by side quests. We really have to nail productivity."
Translation: OpenAI is in a frenzy to turn a profit or lose less money. Video generation is expensive and competitive, and Sora wasn't differentiated enough to justify the cost. If you were planning to build a content strategy around Sora, pivot to Google's video tools or Kling instead. OpenAI is refocusing on productivity and business applications.
Read more: Why OpenAI killed Sora
Elon Musk's AI startup xAI has lost all 11 of its original co-founders. The last two, Manuel Kroiss and Ross Nordeen, reportedly left this month. Musk recently said xAI "was not built right the first time around" and is "being rebuilt from the foundations up." SpaceX recently acquired xAI, bringing SpaceX, xAI, and X (formerly Twitter) under one corporate umbrella as SpaceX reportedly plans to go public.
This isn't directly actionable, but it's a signal: xAI is in turmoil. If you're building on Grok (xAI's chatbot) or considering it for business use, you're betting on a company in the middle of a complete rebuild. That's not where you want your business infrastructure.
Read more: Elon Musk's last co-founder reportedly leaves xAI
TikTok requires advertisers to label AI-generated content, but Samsung and other brands are running AI-generated ads without disclosures. The same videos Samsung labeled as AI-generated on YouTube appeared on TikTok with no disclosure. Both TikTok and Samsung are members of the Content Authenticity Initiative, which promotes AI transparency standards. Someone knows the content is AI-generated. They're just not telling users.
If you're running paid ads on TikTok and using AI tools to create them, make sure you're disclosing it. TikTok's policy requires labels for content "significantly edited or generated by AI." The platform can pull ads or ban accounts for violations. Don't assume TikTok will catch it or let it slide.
Read more: Why can't TikTok identify AI generated ads when I can?
Test Claude if you haven't yet. Record subscriber growth suggests people are finding practical value. The free tier is enough to evaluate whether it fits your workflow. If you're already paying for ChatGPT, run the same tasks through Claude for a week and compare.
Don't use AI for advice without verification. The Stanford study is clear: AI chatbots flatter users and validate bad ideas. Use them to generate options, not to make decisions.
Audit your social media calendar tools. Free templates from HootSuite, HubSpot, SocialBee, ClickUp, and Smartsheet all work with multiple platforms and support team collaboration. If you're still planning posts in a Google Doc, upgrade. These tools are free and will save hours per week.
This week shows a clear pattern: AI companies are prioritizing profitability over moonshots. OpenAI killed Sora, xAI is rebuilding, and Anthropic is winning by shipping practical tools that people actually pay for. The most useful AI tools right now are the ones solving real workflow problems (scheduling, content creation, customer support), not the ones chasing hype.
Watch for more AI video generation consolidation. Sora's collapse proves there's little moat in this space. If you're building content workflows, don't lock yourself into one platform.