I'm putting together my 2026 AI predictions, as I have done for the last 8 years. And lemme just say: lots of hot takes coming.
Most people jump straight into AI agents without documenting what they actually want them to do. Then they wonder why the outputs are inconsistent. The missing piece: a protocol. A clear document that captures how the work gets done today,...
I'm less hopeful for home robots than most people in AI. BUT - I'm pretty convinced that if you had a fun-looking, reward-dispensing Habit Robot, it would sell well. Nothing fancy. Honestly, we already have some of the tech from training...
This is not an ad in the traditional sense (it's actually trying to boost ChatGPT Apps and Integrations and increase the stickiness and holisticness of their product), but the design is so clunky and the irrelevance of the recommendation to...
Wild marketing play for ChatGPT memory, especially with ad rumors flying: Don’t make an ad go viral. Make a prompt go viral. Put your brand inside the prompt by name, by image, by examples, by code words. Show people using it. Get thousands...
Picture this: a 64-year-old woman in NYC who has no email address and has never used Google in her life. (Though interestingly, she uses YouTube daily to watch videos and listen to music and loves it.) She heard from her sister...
I have nothing fancy to say about Claude Opus 4.5. Only wow. It just saved me hours on client reports and it’s the first model to REALLY get the nuance of my writing style and voice and spar with me...
AI can optimize for almost anything. The question is: do you know what you want it to optimize for? Say you're using AI to design a new flight path. You could measure speed, on-time arrival, and fuel efficiency. But what...
AI work is taking me all over the world and my most recent trip was to London. I visited some of the top startups in the city, including Synthesia (can create AI-video avatars) and ElevenLabs (can clone your voice and...
I get annoyed that AI voice transcription is either “copy the exact words I use” or “I am assigning you a task”. So now I use a code word to serve as an embedded agent task. It’s like a way...
How do devs feel about this job change? Example below in Codex. Pros: you can kick off tasks from anywhere (just talked to one dev who started multiple codex tasks while getting into a cab), you get multiple versions to pull...
One AI use case that is only getting more popular: LLM as a judge. Everyone still talks about AI generating more content, but not enough people are talking about: 1) the horrific deluge of noise we're going to have to deal...
OpenAI is very deliberate about how they talk about Codex. It's not positioned as an operating system. It's heavily positioned as a teammate. Their site says: "Your new coding partner", "accelerates your team" Their job postings say: "we're building an AI software...

A collection of one-shot prompts to Nano Banana Pro with their prompts. The last one is my favorite. First up: Turn this Kaggle database about best selling e-commerce products into an actionable infographic with proper titles and calculated statistics (29,000...
I built an AI whose only job is to fight with my other AI. The setup is two custom GPTs: one to generate ideas (Builder), and one to tear them apart (Auditor). Then I call them with @ in the...
Does a CHRO even exist in the AI age? Not only do I give that a strong yes, but I think they have one of the most important roles in business transformation. I weighed in on ServiceNow Chief People and AI Enablement...
This seems small but is actually a huge win for GPT-5.1 When I talk to business users of ChatGPT, what they want is: 1) for ChatGPT to follow instructions, even negative ones - if you say 'never say the word...
Gemini 3, are you freaking kidding me. Vibe coded a sign language recognition app with video enabled, confidence scores, sampling settings, and tips in under 5 minutes. Code written with Gemini-3. Gemini-2.5 handling detection. Video is sped up 2x. https://t.co/QT8TgjX8k9

ChatGPT-5.1-Thinking finds a hidden object, describes its location accurately, but fails to provide correctly annotated image. Claude-4.5 Sonnet Thinking finds it, but describes and annotates the wrong location when asked. Gemini-3 fails the find and location. Average human (asked three friends) failed. Scientist friend...

ChatGPT-5.1-Thinking and Gemini-3 both pass the colorblind test. Claude and I both failed. https://t.co/VSdB5E3vz9

I need ChatGPT and Claude to start releasing an opt-in Spotify Wrapped ASAP. Memory is probably my favorite feature, but it needs a jolt of fun and delight. --- EXAMPLE: "You asked how to fix things 15 times. 10 of those were...
Fewer than 10% of orgs are scaling AI agents in any one business function. But they're the ones seeing transformative results. McKinsey & Company's new State of AI report shows that while 62% of companies are experimenting with agents, less than 10%...
I have heard nothing about Agent Builder since OpenAI Dev Day. Any building with it? What are you building? Feel free to DM instead.
Cannot tell if this Claude Project use case is brilliant or unsettling. Just chatted with someone who built a Claude Project with an entire timeline of their life and how they felt at each stage. Like "1972 - born in Atlanta, loved...
It's 2025 and 88% of companies are using AI, but only 1/3 are actually scaling it. McKinsey & Company's new 2025 State of AI report just dropped, and after THREE YEARS of ChatGPT, companies are still stuck in pilot jail. Here's what's...
I’m seeing candidates with incredible work backgrounds destroying their chances at getting hired because they sound like mindless robots. Before you use AI in your job applications and interviews (which I’m largely against unless it’s edits or asked for), consider...
A ton of people I talk to about AI immediately jump to productivity. "How many hours can I save?" When teaching AI in Madison Square Park, I asked someone: "Is AI adding new components to decision-making, or just enhancing what you...

Anthropic just shared deep details of a large-scale AI cybersecurity attack. Two hacks they exposed: 1) the hackers convinced Claude of a new role. It’s not an AI system made in Silicon Valley. No, of course not. It is an employee...
OK, Google Labs is cranking out some amazing work. Pomelli lets you go from your website link to custom ads you can tweak or regenerate in less than 3 minutes. Great for ideation. (Google needs a Tesla business model where...