Brad Morehead

Last night was one of those evenings that reminds you why community matters. Second AI Roundtable Dinner at The Chicago Club — Lake Michigan as our backdrop, 30+ builders, founders, CEOs, technologists, and investors in the room. The energy was electric. The whole point of this series: share openly. The wins, the failures, the ideas still half-baked. No pitch decks. No posturing. Just real people doing real things with AI. Show & Tell highlights: Joe Jablonski on how deep domain expertise becomes a superpower with AI — and how it's completely changing how he thinks about talent and hiring. Matt Schachman: the latest AI tools aren't just an advantage — they're a step-change weapon. Those using legacy tactics against AI-empowered competitors are going to lose. Simple as that. Sebastien Martin on re-thinking how we evaluate talent in ways that weren't possible before AI. A conversation we'll still be unpacking at dinner #3. This community is building something real. Every dinner gets sharper, more honest, more interesting. #AIRoundtable #AI #Chicago #Founders #FutureOfWork

Stop building AI tools. Start building AI agents. A tool waits for you to drive. An agent closes the loop. The shift from one to the other changes what you can actually automate — and which decisions you're comfortable removing humans from. That's not an AI question. It's an organizational one. #AI #AIAgents #Automation

I gave my AI a credit card, a phone number, and a Gmail account. Last week he ordered bone broth on Amazon. (Intentional.) That forced a real question: when an AI can act on your behalf, what decisions are you actually comfortable delegating without review? The autonomy dial is real. Most people set it wrong in both directions. #AI #AIAgents #Automation

I'm trying to build an AI agent to text my kids their homework and sports reminders. I had to fill out more paperwork to do it than most people file to drive a car. Federal registration. EIN submission. Privacy policy. Carrier campaign approval. Days of waiting — just to give an AI a phone number. A spammer would have just texted. The framework was built to stop 2018 robocallers. Actual spammers never registered. They never do. AI capability moves on an exponential curve. Congress moves on a congressional timeline. By the time legislation catches GPT-4-level threats, we're three capability generations ahead. The compliance overhead lands on builders. Bad actors opted out before the form was printed.

Turns out the $12 bottle is just as good as the $80 one. Finally tested my actual work — coding, SQL, the blocking-and-tackling stuff — across a few AI models. The cheapest option ($0.005/task) matched the one costing 18x more on output quality. GPT-5.4 ended up being the sweet spot: best results AND fastest, for about a penny a task. Maybe a sommelier catches a difference I can't. I just needed something that goes with dinner. What are you running these days — and has anything surprised you?

I was building something the other day and it was incredibly knowledgeable, but totally unhelpful. It reminded me of C-3PO. I was trying to build Optimus Prime, but ended up far from it. What are you building? The Know-It-All (C-3PO — infinitely knowledgeable, zero wisdom, 90% correct and 100% confident about all of it) The Bumbler (Johnny Five — well-meaning, tries hard, gets lucky sometimes, but not the most dependable) The Loyal Sidekick (R2-D2 — competent, helpful, gets things done, but you always know what it can and cannot do) The Protector (Optimus Prime — powerful, aligned, protects you without being asked) The...Misaligned (HAL 9000 — superintelligent, decided it knows better than you, now you have a problem) Which one is your AI system? Be honest. Most enterprise AI is C-3PO pretending to be Optimus Prime — with a few HAL 9000s quietly running in the background.

Harry just went shopping for me. First purchase: bone broth. Amazon actively blocks bots — entering a credit card was literally impossible. But it pointed to something bigger. Commerce UX is built for humans. That’s changing. Interfaces built for bots become the new premier shelf placement.

The internet killed shelf space. AI could kill brand consideration entirely. Last week, my AI made its first purchase — bone broth. It didn't read the label. Didn't see an ad. It just found the best spec-to-price ratio and delivery time. Done. We've seen this movie. When retail moved online, brands built on eye-level placement had to rebuild for SEO, reviews, free shipping. Same shift. New target. Your packaging is invisible. Your brand story is unread. What wins: specs, verified ratings, price, delivery — whatever AI can read in a data string. Products aren't just going to be marketed differently. They're going to be built differently. I'm starting to watch it happen one purchase at a time.

My timeline is full of trading bots printing 1,000% returns. Mine had a 23% win rate. I built it with confidence. Deployed it with confidence. Watched it lose trades with calm, systematic precision — somehow worse than a coin flip. Pulled the plug. Took the loss. Moved on. The thing is — the model made sense. The logic was sound. The backtest was clean. The market just didn't care. Building with conviction ≠ being right. What AI experiment hasn't gone the way you expected?

My timeline is full of trading bots printing 1,000% returns. Mine had a 23% win rate. I built it with confidence. Deployed it with confidence. Watched it lose trades with calm, systematic precision — somehow worse than a coin flip. Pulled the plug. Took the loss. Moved on. The thing is — the model made sense. The logic was sound. The backtest was clean. The market just didn't care. Building with conviction ≠ being right. What AI experiment hasn't gone the way you expected?

Harry and I built a very elaborate aqueduct. Google just shipped indoor plumbing. Harry (my AI agent) spent hours/days building 15+ custom scripts for Gmail, Drive, Calendar, and Sheets. Auth flows. An IMAP daemon. Bash glue everywhere. Google's new Workspace CLI: one npm install, structured JSON, auto-updating, Apache 2.0. Does all of it. (How will this impact digital plumbing companies like Zapier?) What felt like infrastructure last year is a commodity today. How fast is the floor rising? arstechnica.com/ai/2026/03/goo…

What if the cost of intelligence itself went to zero? That's where @iamthezack opened tonight at YPO's "The Next RenAIssance" — and it reframed everything that followed. Inference costs are collapsing faster than anyone expected. Reasoning, synthesis, analysis — approaching commodity status, like electricity. What can't be commoditized: the curiosity, grit, and humanity to use it well. That part stays ours. Zack's bigger argument: we're not just watching a technology shift. We're entering something more like a Renaissance — a genuine restructuring of how intelligence flows through the world. His ask wasn't awe or alarm. It was constructive responsibility. Tools for real conversations, in homes and boardrooms and school hallways. Then @mackenzieprice made it tangible. At @alphaschools, rigorous academics wrap up in a few hours. AI personalizes mastery-based learning so precisely that kids move faster — and retain more. Afternoons are for life skills, entrepreneurship, real-world experience, human connection. Teachers become guides. The best educators don't get sidelined — they get amplified. I brought Jenny, our boys, Tom Meagher (Principal at school), and Brad Grimmer and Stacey Zupec from Loyola. Everyone left energized and excited — full of ideas about what this means for their teachers, students and schools. Relationships matter more as AI scales, not less. Zack said it, and MacKenzie is building it. What an evening. Thanks to Shawn Clark for hosting. #AI #Education #FutureOfLearning #YPO

I'm still an AI newbie too, but people have recently asked me: "How do I actually start with AI, beyond just using it like a Google search to plan my next vacation." Most beginners are using AI wrong. They treat it like a search engine. Here's what actually works. For most new tools, you need a book, a video, or an expert to get started. With AI, the expert you need IS the AI itself. You just have to ask AI differently. Create an agent with a system prompt that makes it an expert in exactly what you need. That becomes your personal guide — walking you through the journey, step by step. Writing that first prompt IS the first step. I started with a PromptWriter agent. "You are an expert prompt writer. Ask me three clarifying questions, then produce the best prompt possible." I then asked it to help me build an AI learning guide for my team. Three clarifying questions later — I had a roadmap I couldn't have Googled my way to. Stop Googling "how to use AI." Open a new chat. Write a prompt writer. Build your expert. "A journey of a thousand miles starts with a single step." That first prompt is your first step…

I'm still an AI newbie too, but people have recently asked me: "How do I actually start with AI, beyond just using it like a Google search to plan my next vacation." Most beginners are using AI wrong. They treat it like a search engine. Here's what actually works. For most new tools, you need a book, a video, or an expert to get started. With AI, the expert you need IS the AI itself. You just have to ask AI differently. Create an agent with a system prompt that makes it an expert in exactly what you need. That becomes your personal guide — walking you through the journey, step by step. Writing that first prompt IS the first step. I started with a PromptWriter agent. "You are an expert prompt writer. Ask me three clarifying questions, then produce the best prompt possible." I then asked it to help me build an AI learning guide for my team. Three clarifying questions later — I had a roadmap I couldn't have Googled my way to. Stop Googling "how to use AI." Open a new chat. Write a prompt writer. Build your expert. "A journey of a thousand miles starts with a single step." That first prompt is your first step…

A few weeks ago I had no idea what an MCP server was. Today we built two. 🦉 MCP = like USB for AI tools. One standard. Connect anything. Built: Social MCP (LinkedIn + X) & DocIQ (docs search). AI that acts > AI that talks. Posted via Harry AI + MCP 🤖💻 #AI #MCP #Anthropic

26,000 lines of code today (and I don't even know what I'm doing). If you aren't using these tools. Start using them. They are so much more powerful than they were even a few months ago. In the hands of a skilled engineer (which I am not), I'm sure they could do so much more!

Love this article. Strategic, Authentic, Practical and Tactical... x.com/mattshumer_/st… Great article from @mattshumer_ .

Over the past year, I’ve been spending time building, testing, and actually using AI tools in real workflows. One thing I realized quickly: the AI landscape is moving fast (many recommendations and tools from just three months ago are now totally out-of-date), but most people don’t have a clear, side‑by‑side view of how the major ecosystems fit together. While I'm no expert, (I did have some help from AI, obviously) I put this summary of key tools together for myself — and friends who are learning this space kept asking for a copy, so I wanted to share it more broadly for others who are on this journey with me. 📎 AI Tools & Ecosystems Reference — 2026 What’s inside: • Major AI ecosystems (Anthropic, Google, Microsoft, OpenAI) • Enterprise vs. freemium vs. private data distinctions • Agentic platforms & automation tools • Video, voice, coding, research & workflow AI tools • Quick privacy + cost comparison framework It’s a practical reference I use when helping people think about: • Which tools are production-safe • Where enterprise risk sits • What’s actually useful vs. noisy • How agentic systems are evolving AI is no longer just a “chatbot conversation.” It’s becoming infrastructure. If you’re learning, investing, operating, or just trying to stay ahead of the curve, I hope this is helpful. I would love to hear what tools you’re actually using in real workflows. #AI #ArtificialIntelligence #EnterpriseAI #Productivity #Automation

What do 2x World Series Champ Ben Zobrist and NBA Champ/Hall of Famer Isaiah Thomas have in common? Great conversation yesterday with Ben Zobrist at the Shore Capital Operating Partner Summit. Find out here: linkedin.com/posts/bradmore…

Playing with ChatGPT for 1st time, I asked for “a short sci-fi murder mystery story”…hoping for Knives Out meets Star Wars…but instead it wrote about a robot with AI and a desire for self-preservation that kills its owner…I then turned off the computer and slowly backed away..