AI: Pretending, Building, Stalling
· The Fluency Briefing
The Fluency Briefing
Your Guide to What's Happening in AI and Why It Matters to You
Tuesday, April 21, 2026
Three stories landed on my desk this Tuesday that, taken together, tell you exactly where AI stands right now: a med student in India built a fake MAGA influencer and milked thousands from lonely men, an AI can now identify anonymous writers from a single paragraph, and 95% of enterprise AI pilots are still quietly failing. The through-line?
AI is fantastic at pretending - pretending to be a person, pretending to know who you are, pretending to transform your business - and we're only starting to figure out which pretenses actually hold up.
Today in AI:
- Fake MAGA Girl, Real Money - A broke med student in India used Google Gemini to create a fictional conservative influencer named Emily Hart, then sold bikini photos to what he calls "super dumb" men for thousands of dollars. Gemini reportedly called the conservative niche a "cheat code." Wired
- Your Anonymous Posts Aren't Anonymous Anymore - Anthropic's Opus 4.7 can now identify writers from short text snippets using stylometry - no custom code needed, just a prompt. What took researchers months in 2023 now takes a chatbot seconds. Lesswrong
- Codex Hits 4 Million Developers a Week - OpenAI says Codex jumped from 3 million to 4 million weekly users in just two weeks and is launching Codex Labs to embed its engineers inside enterprise orgs. Virgin Atlantic, Cisco, and Rakuten are already using it across real workflows. OpenAI
- Tim Cook Steps Down, Hardware Guy Takes Over - Apple named 25-year veteran John Ternus as its next CEO, effective September 1. Ternus reportedly opposed the Vision Pro and the scrapped car project, and he's already reorganized hardware engineering around an AI platform. BBC
- Yelp Becomes Your AI Concierge - Yelp rolled out a major upgrade to its AI Assistant, adding direct booking through DoorDash, Zocdoc, and Vagaro. The chatbot now gets its own tab in the app and explains why it recommends each result by pulling from reviews. The Verge
- 95% of Enterprise AI Pilots Are Failing - A widely cited MIT-backed analysis confirms what many suspected: almost all enterprise generative AI projects stall before production. The problem isn't adoption - it's that LLMs produce language, but companies run on memory, context, and constraints. Fast Company
- Fermi's Leadership Implodes - Rick Perry's AI power startup Fermi lost both its CEO and CFO within 48 hours. The company IPO'd at a $16 billion market cap last October with zero revenue and zero customers - it's now worth $3.4 billion and still has no anchor tenant. Fortune
- Stripe Alumni Raise for AI Fintech - Seapoint, an AI-powered financial operations platform built by former Stripe engineers, raised a 7.5 million euro seed round to scale across Europe. The startup targets the messy back-office finance work that most AI tools ignore. Sifted
Today's Takeaway:
Here's the thing about AI's current moment: the tools that work best are the ones that know their lane. OpenAI's Codex is thriving because it does something specific - write, review, and debug code - inside workflows developers already use. That's why it jumped a million weekly users in two weeks, according to OpenAI. Meanwhile, as Fast Company reports, 95% of broader enterprise AI pilots are failing because companies expected a chatbot to magically understand their entire operation. Translation: asking an LLM to "transform the business" is like hiring a brilliant poet to run your supply chain.
Yelp's upgrade tells the same story from the consumer side. Instead of building a general-purpose AI, Yelp pointed its assistant at a narrow, high-value task: matching you with a local business and letting you book it on the spot, as covered by The Verge. That's not glamorous, but it's useful - and useful is what survives. The pattern is clear: AI succeeds when it's scoped tightly and plugged into real decisions. The companies treating it like fairy dust are the ones writing post-mortems.
💡 Fluency Moment - Building your AI fluency, one term at a time.
"Stylometry"
In plain English: AI analyzes writing patterns to identify who wrote something, even anonymously.
Think of it like: Like a fingerprint, but for your word choices, sentence rhythm, and punctuation habits.
Why you'll hear about it: Anthropic's Claude can now de-anonymize writers from a single paragraph in seconds.
🧰 Your Toolkit
5-Minute Quickstart: Spotting AI-Generated Content in the Wild
- Open any social media feed and look for profile photos with unusually perfect skin, symmetrical features, or blurry ear and hair edges.
- Check the background of suspicious images for odd details like melting objects, extra fingers, or text that looks like gibberish.
- Search the profile or post for inconsistencies - does the person's story change, or do all photos look like they were taken in the same lighting?
- Ask ChatGPT: 'What are 5 quick signs that a photo of a person might be AI-generated?' and save the list on your phone.
- Use a free tool like Google Reverse Image Search - drag any suspicious photo into images.google.com to see where else it appears online.
- Before trusting or sharing content, ask yourself: 'Does this person want something from me?' Scammers use AI personas to build fake trust fast.
Next, explore how to use AI tools like ChatGPT to help you fact-check headlines or suspicious messages before you share them. Learning to question what you see online is one of the most valuable skills in the AI age.
The Bottom Line
The Pattern: The gap between AI that performs and AI that pretends is getting wider. Tightly scoped tools like Codex and Yelp Assistant are delivering real results, while broad "let AI run things" initiatives keep crashing into reality - and that includes billion-dollar infrastructure bets like Fermi.
Why It Matters: If you're evaluating AI for your team or your life, the lesson is the same whether you're a Fortune 500 exec or someone trying not to get catfished by a fake MAGA nurse: specificity wins, hype loses. The companies and individuals who ask "what exact problem does this solve?" will pull ahead of those chasing the abstract promise of intelligence.
Your Move: Pick one repetitive task you did this week - just one - and test whether an existing AI tool can handle it end-to-end. Not "transform your workflow." Just one task. That's where the 5% of successful pilots started.
What We're Working On
✨ Founding Cohort Special - 60% Off! - Use code MAF20 to join for just $20/month (regularly $50). Get weekly group sessions & workshops, self-paced courses for all levels, access to tools & templates, challenges with peer feedback, and 24/7 support community. → Join Now
✨ Free 30-Minute AI Consultation - Discover how My AI Fluency can help your business unlock the potential of AI. We'll discuss your goals, explore practical AI opportunities for your industry, and outline clear next steps. → Schedule Free Call
✨ How AI-Fluent Are You? - Test your AI fluency with our interactive quiz. See how you stack up and discover what to learn next. → Take the Quiz
💬 Community | 📞 Book a Consultation | 🌐 Website
Fluently yours, The My AI Fluency Team