Hello WKND AI Warriors!

Deepfakes are getting so convincing that experts say you need verification habits like safe words and second-channel confirmations, not just spotting visual glitches.

Also, Microsoft’s CEO warns AI will lose public and political support if it cannot prove clear real-world value worth the electricity it consumes.

Plus, MIT’s new recursive approach lets language models work across roughly 10 million tokens by generating code to search and slice massive documents.

Oh yeah, and Google’s AI Mode can now opt-in personalize answers using your Gmail and Photos, while Gemini adds free SAT practice tests with The Princeton Review.

Today’s newsletter includes:

• 🎓 AI COURSE OF THE WEEK

• 📰 AI NEWS RECAP

• 🤿 AI DEEP DIVE

• 📝 AI PROMPT OF THE WEEK

• 🎨 AI IMAGE OF THE WEEK

🎓 AI COURSE OF THE WEEK

Learn more about AI in 1 hour than you did all of last year.

Get instant access here.

Oh yeah…its FREE!

📰 AI NEWS RECAP

How To Spot AI Deepfakes In 2026👀 

Deepfakes used to be a party trick.

Now they are an operational risk.

I read a guide on spotting deepfakes in 2026 and the main takeaway is uncomfortable: your eyes are no longer a reliable security system.

What works now is not better intuition. It is better habits.

What to watch for

1. Emotional pressure
   If a clip is designed to make you panic, feel outraged, or act fast, assume that is the point. Deepfakes are persuasion tools first.

2. Movement and physics
   Faces are getting near perfect. Bodies, lighting, and motion still betray the edit. Look for weird head movement, inconsistent shadows, unnatural blinking, or edges that shimmer when someone turns.

3. Audio that feels right but wrong
   Voice clones are good enough to fool you, especially on a quick listen. The tell is often cadence, breath, timing, and phrasing that does not match the person you know.

What to do instead of guessing

1. Confirm through a second channel
   If someone asks for money, access, or urgency, verify through a separate method. A call back. A known email thread. A direct message to a saved number.

2. Use a simple safe phrase
   One sentence your family or team can use to prove it is really them. Boring, effective, fast.

3. Make verification a process, not a vibe
   Deepfakes are the next phishing. The fix is not becoming a detective. The fix is building a protocol so people know what to do when something feels off.

The real shift is this:

In 2026, the question is not “Can you spot a deepfake?”
It is “Do you have a routine that prevents one from hurting you?”

Microsoft’s CEO warns that if AI does not produce clearly useful real-world outcomes, the industry could lose public and political support to keep consuming vast amounts of electricity. He argues future investment and regulation will hinge on demonstrable societal value, not just impressive benchmarks.

MIT’s new “recursive” framework lets language models handle inputs around 10 million tokens by generating and running code to search and slice huge documents. This enables whole-codebase reviews or long legal analyses without fitting everything into a single context window.

Google’s Gemini app now includes free full-length SAT practice tests built with The Princeton Review, offering instant scoring and explanations. It also generates personalized study plans so students can focus on their weakest topics.

Anthropic introduced a new written “constitution” for Claude that more explicitly commits it to protect humans and avoid harmful power-seeking behavior. The goal is to bake clearer, stable safety principles into Claude’s behavior as the model gets more capable.

Google’s AI Mode in Search can now, with opt-in, use your Gmail and Photos to personalize answers. It can, for example, suggest travel plans that fit your past trips and brands or tailor shopping advice using your upcoming flight details.

Google has reportedly hired the team behind Hume AI, a startup focused on emotionally expressive AI voices. The move signals a stronger push to make Google’s assistants sound more natural, nuanced, and human-like in conversation.

🤿 AI DEEP DIVE

Wish AI was your coworker?

Claude ‘Co-Work’ is here.

Why wait?

📝 AI PROMPT OF THE WEEK

Copy and paste this into your favorite chatbot.

USE CASE: Research brief with sources and gaps.

PERSONA: Analyst.

CONTEXT: Topic is [topic], audience is [audience], deadline is [date].

TASK: Create a research brief with key questions, source list, and gaps to close.

FORMAT: Objective, Questions to Answer, Key Terms, Sources table, Gaps and Close Plan, Five Key Takeaways.

TONE: Neutral, scholarly.

🎨 AI IMAGE OF THE WEEK

Give AI image creation a try for yourself.

Copy the text below into Google’s AI image creator!

A large, spacious living room interior in vibrant Pop Art aesthetic with contemporary Japanese-inspired playful florals. The scene is overloaded with bold graphic patterns--oversized florals, polka dots, stripes, halftone textures, and comic-style shapes--covering every textile surface: curtains, rugs, cushions, wallpaper, and sofa upholstery. At the center of the room stands a Porsche 911 as the main focal subject, shown in a clean 3/4 view, positioned like a sculptural centerpiece on a low platform or directly on the rug. The Porsche’s paint is rendered in flat, vivid color blocks with pop-style decals and patterned reflections that echo the room’s motifs (florals, dots, stripes), but the car remains clearly readable and iconic. Modern retro-inspired furniture surrounds the car: low modular sofas, geometric coffee tables, pop decorations, posters, and graphic color-block panels on the walls. Add comic speech-bubble elements and playful pop signage integrated into the decor (no readable brand text). Include one female figure in a mid-shot seated on the sofa, faceless, with visible hair and expressive posture, wearing a floral dress that matches and blends with the room’s patterns. Flat, vivid colors, low-contrast even lighting, minimal shadows, crisp vector-like edges, surreal playful atmosphere, highly detailed pattern layering, wide-angle interior composition.

How can you help?

Refer my newsletter to help others learn AI.

Missed last week’s edition?

Find all of my newsletters here!