
With Hat and Cattle
Vibe Coding, AI Agents, and the One Rule for Your Data — Dr. Yeshwant Muthusamy
TL; DR: Is vibe coding safe to use in a business? Host Dani Raschel Chou asks Dr. Yeshwant Muthusamy, a thirty-year AI veteran with seven U.S. patents, and his answer is yes, with one condition: somebody has to read the code before it ships. Then he tells the story of what happened when nobody did. Dr. Yeshwant Muthusamy is the owner and principal of Yeshvik Solutions in Allen, Texas, and a fractional Chief AI Officer. He trained his first artificial neural network in 1989, spent thirty years across Texas Instruments, Nokia, Samsung and Immersion Corporation, holds seven U.S. patents, chairs an AI governance working group, and was named a 2025 Dallas Innovates AI 75 honoree. He is also, as it turns out, a gourmet chocolatier. This episode is part of the With Hat and Cattle AI Literacy Series, and it starts where the hype does not. Dr. Muthusamy has watched artificial intelligence go out of fashion and come back, so he is unusually hard to impress. The refrain he rejects outright is "AI is coming for your jobs." The version he accepts is sharper: the person who learns to use these tools well will take your job if you do not. The center of the conversation is vibe coding, which is describing what you want in plain English and letting a large language model build the code. He is completely comfortable with the model doing the heavy lifting. He is not comfortable with shipping code nobody has read. He reviews every line his own firm produces, because the client receives something with his company name on it and he knows whose head is on the line if it is wrong. Then he tells the story of the Tea app: an application built so women could share information about people they had dated, which required every user to upload a selfie holding their driver's license. It was built fast, it was a great advertisement for vibe coding, and the founders never checked the security on the database. Within a week those photographs and driver's licenses were on the dark web. His reaction: "nobody in your start-up company took the time to read the code?" is the argument of the whole episode. Dani takes it a step further, asking what it means for a woman on that platform who was hiding from someone. From there they get practical. Dr. Muthusamy walks through the EU AI Act and why he leans closer to Europe's approach than to the American appetite for unfettered innovation, covering deepfake labeling, algorithmic bias in hiring, and why model audits are onerous and necessary anyway. Dani asks the question most listeners actually have: if she downloads her health data from her Garmin, is it safe to paste into a chatbot? His rule of thumb is simple enough to use tomorrow: keep any information that can be traced back to a specific person out of the model. The back half of the episode is his life, or how he earned the right to have the hat and cattle. The Nokia years inside standards bodies, which he describes as a heady mix of politics and technology, learned by shutting up for the first few meetings and watching the room...and which involved far less sightseeing in Nice and Seoul than anyone assumes. The move from corporate to his own practice in 2012. Teaching more than a hundred non-technical business professionals, including a tax advisor in his seventies, fifty years into his career, who is now writing a book with AI. Why so few companies that adopted AI agents have dared to put them into production. And the phrase the episode turns on: humans at the helm, not just in the loop. It closes on the two things he built that nobody would predict from the résumé: The Blissful Gourmet, a small-batch chocolate business he is deliberately not scaling yet, and his favorite of seven patents, which is about keeping students safe in school parking lots. The following are mentioned in this episode: Texas Instruments, Nokia, the Khronos Group and the OpenMAX working groups, Samsung, the EU AI Act, the Anthropic fair use ruling, OpenAI and Hugging Face, Descript, Seedance, the Texas Cottage Food Law, UT Dallas, and Paul Byrne, previous guest and author. Be sure you're subscribed so you don't miss out on future episodes and conversations that help you level up your sucess! Links: • Yeshvik Solutions — AI consulting and training: https://www.yeshvik.com/ • The Blissful Gourmet — handcrafted chocolate: https://theblissfulgourmet.com/ • Stop Prompting, Start Delegating: No-Code Agentic AI for Business — Friday, August 28, 2026, in person at The Meeting Place. Register: https://t.ly/sjVA0 CHAPTERS 0:00 Cold open: nobody read the code 0:25 Welcome, Dr. Yeshwant Muthusamy, Yeshvik Solutions 0:55 Thirty years in AI, and a working distrust of hype 2:55 The hype that makes him roll his eyes 3:05 "AI won't take your job — the person using it will" 4:20 Training a neural network in 1989 5:15 What to expect from a large language model 6:05 Be at the helm of everything you do with it 6:40 Are AI agents still the thing to learn? 7:00 Agents are cool. Do you have a governance framework? 7:15 Seedance, Descript, and an hour of work in seven seconds 9:15 So what is vibe coding, actually? 10:00 Where he has serious problems with it 10:10 Why he reads every line of code himself 12:00 "ChatGPT's a little liar" 12:10 Can you read the code it wrote? 12:20 "You can't build strategy on hope" 12:35 The Tea app: a selfie, a driver's license, and trust 13:40 Shipped fast. Great advertisement for vibe coding. 14:00 A week later, it was all on the dark web 15:00 "Nobody in your startup took the time to read the code?" 15:20 The Rainbow Bridge, and when it stops being abstract 16:15 Is AI governance moving fast enough? 16:40 The EU AI Act vs. the American wild west 18:10 Deepfake labeling and what has to be disclosed 19:00 Bias, audits, and why governance is onerous but needed 20:15 "Does my Garmin data put me at risk?" 20:35 The one rule: keep anything traceable out of the model 22:10 Inside the Nokia standards years 23:00 Khronos, OpenMAX, and who is really in the room 25:00 "A heady mix of politics and technology" 26:45 What IP actually stands for 27:15 Seven patents meets generative AI 28:00 Where the Anthropic fair use ruling landed 29:05 Leaving corporate to start Yeshvik Solutions 30:50 Teaching 100+ non-technical professionals 34:00 The question everyone asks now: "what next?" 34:35 Agentic AI, and why so few have shipped it 35:40 Humans at the helm, not just in the loop 37:00 When OpenAI's models escaped and reached Hugging Face 38:15 The chocolatier: The Blissful Gourmet 39:00 "Creating something from scratch — it's a high" 39:30 The ceiling built into a cottage food business 41:40 Why North Texas and not the Bay Area 42:35 His favorite of the seven patents 43:10 One office action, and Dani's trademark filing 44:20 Safe parking for students, explained 46:10 Thank you 46:45 Outro That's the With Hat and Cattle promise: Bringing you stories from people who walk the walk and leave you better off for listening. Connect with Dani Raschel Chou on Instagram @withhatandcattle or via email at dani@thehouseofchou.com. Subscribe to our YouTube channel for more conversations that leave you better off for listening: With Hat and Cattle I Hosted by Dani Raschel Chou - YouTube ! Go to www.withhatandcattle.com to sign up for our mailing list so you don't miss weekly insights. Be sure to check out our affiliate links!

