Disrupt or Defend

How AI is disrupting the restaurant experience | Ep. 28

Spotify and Netflix both know what you like. The one place that exists for your literal taste, the restaurant, hands every guest the same printed card. ㅤ Daniel Kazani [https://www.linkedin.com/in/danielkazani/] sits down with Amir Khoshbakht [https://www.linkedin.com/in/khoshbakht/], co-founder and CTO of Localia, to talk through what happens when AI moves out of the dashboard and into the dinner table. ㅤ Amir's team is rebuilding the restaurant menu as a personalized surface. Items sort themselves to each guest based on prior orders and cross-venue preferences. An in-menu AI assistant answers questions in any language. A new agent on the manager side predicts which ingredients to push or pull back this week. ㅤ Early data from Localia's customers: 15 to 20% more engagement with menu items, and bigger checks per guest. The conversation also covers why most restaurant owners almost hate AI on first pitch, and where AI actually earns its keep when it leaves the dashboard. ㅤ 👤 Guest Bio Amir Khoshbakht [https://www.linkedin.com/in/khoshbakht/] is co-founder and CTO of Localia. He spent over a decade in software engineering, with prior full-stack and blockchain work at Soar Robotics and a co-founder role at Aikido. He started Localia after deciding AI should move out of dashboards and into the parts of daily life people physically experience, beginning with restaurants and cafes. ㅤ 📌 What We Cover * Why every restaurant menu is the last static surface in your customer experience * How Localia's AI Waiter inside the menu changes the order flow * The 15 to 20% lift in engagement when the menu sorts itself around each guest * Cross-venue preference matching: liking the chocolate cake at restaurant A changes what restaurant B shows you first * What the data shows about why low-selling items actually sell low (rarely about quality) * "The Expert," Localia's new agent that predicts ingredient demand for the kitchen * Why most restaurant owners almost hate AI when you walk in pitching it * Where Amir sees hospitality landing: an operating system with fewer tools, not more ㅤ 🔗 Resources Mentioned * Localia [https://localia.ai] * Amir Khoshbakht on LinkedIn [https://www.linkedin.com/in/khoshbakht/] * Spotify * Netflix

How Fractional CTOs use AI to help Startups | Ep. 26

The hype says AI is coming for your engineers. John McKinney [https://www.linkedin.com/in/jsmckinney/], a fractional CTO who has run engineering at AOL and a string of CTO seats since, says it is closer to "super Google." ㅤ Daniel Kazani [https://www.linkedin.com/in/danielkazani/], co-founder of Softup [https://www.softup.co], sits down with John to talk about what has actually changed in how software gets built. The conversation lands on the shifts that matter most for founders and engineering leaders right now: product-minded founders shipping working prototypes with Lovable [https://lovable.dev] before an engineer joins, the dev role looking more like a mechanic and less like an architect, and the cost of trusting Claude Code [https://www.anthropic.com/claude-code] with HIPAA-bound systems. ㅤ John is a 20-year engineer and longtime skeptic of every new wave that hits the industry. He is also a working bassist in two New York metal bands, which says something about the kind of CTO he is. They get into vibe-coded MVPs, why QA engineering might be the role most reshaped by AI right now, the broken state of remote technical hiring, and the difference between sprinting to a real horizon and selling a horizon you can't deliver. ㅤ 👤 Guest Bio John McKinney [https://www.linkedin.com/in/jsmckinney/] is the founder of Merge Conflict [https://www.mergeconflict.io], a fractional CTO and technology-strategy consultancy he started in 2017 in New York City. He came up as a web developer in the Ruby on Rails era, co-founded the agency Ashe Avenue in 2007, and stayed on as VP of Engineering for AOL Core Products after AOL acquired the company. Since then he has held CTO seats at Netcapital, LaterPay, Heyday, and Food52. Off-keyboard he plays bass in the New York metal bands Woe and Glorious Depravity. ㅤ 📌 What We Cover * How Claude Code went from novel to "a staple of every engineer's toolbox," and why John doesn't believe it turns one-x developers into ten-x. * The shift in product roles: product leads now ship working proof-of-concepts before an engineer touches the code, and what that means for PRDs. * A real client story of going from a multi-million-dollar agency build to shipping the same retail and clinician software with Lovable. * The new shape of the dev job: bulletproofing, production hardening, and compliance checking after the founder ships. * Why a HIPAA leak is still your fault, not Claude's, and how exposed API keys on the front end happen. * The poisoned hiring funnel: AI voice modulators in phone screens, 400 unqualified applications in the first hour, and Voight-Kampff-style interviewing for replicants. * What CS 101 might look like when "Hello, World" gets replaced by prompt-writing, and whether junior-to-mid acceleration actually speeds up. * John's read on the "software engineering is solved" claim, the Klarna walk-back, and why the snake-oil crowd from Web3 is the same crowd selling AI now. ㅤ 🔗 Resources Mentioned Merge Conflict [https://www.mergeconflict.io] - John's fractional CTO practice Claude Code [https://www.anthropic.com/claude-code] (Anthropic) Lovable [https://lovable.dev] ChatGPT [https://chatgpt.com] (OpenAI) Grok [https://grok.com] (xAI) VS Code [https://code.visualstudio.com] Selenium [https://www.selenium.dev] Stripe [https://stripe.com] Ruby on Rails [https://rubyonrails.org] Klarna [https://www.klarna.com] (the AI-layoffs walk-back John references) Artisan [https://www.artisan.co] (the "replace your human workforce" billboard campaign) Icon [https://icon.com] (the AI ad-maker Daniel tested) Sam Altman, Ray Kurzweil, the Voight-Kampff test from Blade Runner

How AI is impacting Wealth Management | Ep. 25

Most people think of AI as a productivity tool. Dr. Adam Link [https://www.linkedin.com/in/adamlink/] sees it as something else entirely: a wealth transfer mechanism, one that's quietly shifting money from people who don't understand AI toward people who do. In this episode, Daniel Kazani [https://www.linkedin.com/in/danielkazani/] of Softup [https://www.softup.co] and Dr. Link break down what that actually means for founders, developers, and anyone trying to build financial security in a world where AI agents are replacing entry-level work. ㅤ The conversation spans a lot of ground: the collapse of the college-to-job pipeline, why non-technical entrepreneurs are outpacing developers in AI adoption, the potter's wheel vs. the factory floor, and why the court system is just the debugging process for laws. Concrete, candid, and a bit unsettling in the best possible way. ㅤ 👤 Guest Bio Dr. Adam Link, CFP® [https://www.linkedin.com/in/adamlink/] is the founder of Fireweed Capital, a fee-only, fiduciary wealth management firm based in northern Minnesota. He holds a Doctorate in Computer Science, a CFP® designation, and currently serves as Senior Engineering Manager at Coinbase, where he leads the Cloud Center of Excellence and FinOps teams. Before Coinbase, he worked across multiple tech startups with three exits and an IPO under his belt. He now specializes in active risk management for tech families - and builds his own CRMs over the weekend. ㅤ 📌 What We Cover * Why AI is accelerating the shift from a labor economy to a capital economy - and what that means for how you think about your "war chest" * The college-to-job pipeline breaking: why large tech companies are pausing entry-level hiring and what happens to the path from degree to wealth * The CRM story: how Dr. Link got blocked by a vendor on a Friday and had a fully vibe-coded, custom-built replacement running by Monday morning * AI as wealth transfer - not just to hyperscalers, but to small businesses that cut OpEx, kill vendor dependencies, and keep more of their own money * Vendor concentration risk: when you replace your whole team with OpenAI, you've just created a two-vendor company * The railroad analogy: why AI infrastructure is different from past wealth transfers - and why this time, everyone gets their own train * The franchise model: one friend's idea to build AI-powered websites and license the software out instead of building a SaaS * Why non-technical entrepreneurs are outpacing developers in AI adoption - and why that's terrifying if you're a developer * The potter's wheel vs. the factory: how AI is disrupting the creative flow state that made software engineering meaningful * AI hallucinations in financial advising, why freeform text has no unit tests, and why good professionals still matter ㅤ 🔗 Resources Mentioned * Fireweed Capital [http://fireweedcapital.com] - Dr. Adam Link's wealth management firm * adam@fireweedcapital.com - Contact Dr. Adam Link directly

AI Adoption among German SMEs and Enterprises | Ep. 24

Less than 20% of Germany's "hidden champions" use AI systematically. That single stat sets the tone for this conversation. Daniel Kazani [https://www.linkedin.com/in/danielkazani/] of Softup [https://www.softup.co] sits down with Dr. Annette Doms [https://www.linkedin.com/in/annettedoms/] - EU AI Act strategist, serial founder, and one of the most influential voices on AI in the German-speaking world - to ask the question that most Mittelstand leaders are afraid to answer out loud: are we actually ready? ㅤ What comes out of the conversation is honest and specific. German SMEs aren't failing because they lack data or domain expertise - they have more of both than almost any competitor. The problem is the gap between potential and practice. Too much hesitation, too little experimentation, and a regulatory environment that adds uncertainty faster than it provides clarity. ㅤ 👤 Guest Bio Dr. Annette Doms [https://www.linkedin.com/in/annettedoms/] is a Munich-based AI strategist, author, and serial founder with a PhD in art history and over 13 years at the intersection of technology and business transformation. She is CEO of ICAA Strategists GmbH [https://icaa.io], Vice President at the Bundesverband für KI-Transformation e.V., and Founding Partner of MindMash - a Munich community for forward-thinking technologists meeting every second Tuesday. Named a Top 50 Most Influential Woman in AI in 2026, she specializes in helping Mittelstand companies and executive boards build AI readiness strategies, including EU AI Act compliance. Her book "Von Gutenberg zu ChatGPT" frames the current AI era as a civilizational shift - not a product cycle. ㅤ 📌 What We Cover * Why the German Mittelstand is "AI curious" but not AI ready - and why that gap is dangerous when competitors are moving fast * The "Rhinoceros" concept from Dr. Doms's book: how Mittelstand companies are built for durability, not unicorn exits - and what that means for AI adoption * Why AI readiness starts with clean, structured data in ERP systems - long before any tool gets selected * The step-by-step roadmap for a first AI implementation: assessment, bottleneck identification, scoped pilot, proof of concept, then scale * How to think about AI as part of a company's DNA - not an add-on - and why change must be led from the top * The data sovereignty problem: why Mittelstand companies are hesitant to train on US or Chinese models, and what federated learning and platforms like Catena-X offer as alternatives * The EU AI Act's current ambiguity: how unclear penalty structures and fragmented European infrastructure are adding hesitation, not confidence * Where the next three years lead: the companies that operationalize AI now will set the pace for everyone else ㅤ 🔗 Resources Mentioned * Dr. Annette Doms on LinkedIn [https://www.linkedin.com/in/annettedoms/] * ICAA Strategists GmbH [https://icaa.io] * MindMash [https://www.mindmash.club] - community meetup, every second Tuesday in Munich * Catena-X [https://catena-x.net] - decentralized automotive data ecosystem for sovereign AI infrastructure * Von Gutenberg zu ChatGPT - Dr. Doms's book on AI as civilizational shift (referenced in episode) * Softup Technologies GmbH [https://www.softup.co]

Building internal tools with AI | Ep. 23

Most internal tools are outdated before they ever ship. That's the problem at the heart of this conversation - and it's one that almost every technical founder has felt personally. Daniel Kazani [https://www.linkedin.com/in/danielkazani/] sits down with Dario Di Carlo [https://www.linkedin.com/in/dicarlodario/], CEO and co-founder of bricks.sh [https://bricks.sh], to talk about why internal tooling is quietly draining engineering teams, and why the vibe-coding shortcut will make it worse before it makes it better. ㅤ Dario built bricks.sh after living the problem firsthand: spending months building admin panels at a previous startup, only to ship something already outdated. The platform auto-generates admin panels directly from your database and keeps them in sync as your schema changes - so your engineers stop rebuilding the same tables, forms, and dashboards over and over again. Softup [https://www.softup.co] has seen this same pain up close with dozens of clients, which makes this one of the more honest and grounded conversations on the show. ㅤ 👤 Guest Bio Dario Di Carlo [https://www.linkedin.com/in/dicarlodario/] is CEO and co-founder of bricks.sh [https://bricks.sh], a Milan-based startup automating the creation and ongoing maintenance of internal admin panels. Before bricks.sh, he was a Product Manager at Nibol - Italy's leading workplace management platform - and spent several years exploring the intersection of AI and APIs. He holds an MSc in Innovation Management from Scuola Superiore Sant'Anna and co-founded bricks.sh with CTO Giuliano Torregrossa. The company raised a €1.6M pre-seed round in early 2026 and has signed up 500+ users across 40+ countries. ㅤ 📌 What We Cover * Why internal tools are the most-built, least-appreciated part of any product org - and why that gap creates real friction between engineering and business teams * The difference between dashboards (read-only data access) and operational admin panels (where you actually update users, flag transactions, process refunds) - and why bricks.sh aims to cover both * Why vibe-coding your internal tools with Lovable or Claude Code creates a short-term win and a long-term maintenance tax * The guardrails argument: why giving business users free customization without engineering guardrails turns into shadow IT and shadow AI - fast * How bricks.sh chose startups and scale-ups over enterprise design partnerships early on, and why that bet was the right one for product-market clarity * The bricks.sh product thesis: 90% of internal tools across finance, healthcare, logistics, and ed tech look identical - speed of development will beat freedom of customization * What the future of internal tools looks like: conversational interfaces, MCP-powered data unification, and a single source of truth across Stripe, Intercom, Linear, and Postgres * The founder side: zooming in vs. zooming out, defending a thesis under pressure, and why emotional stability - not enthusiasm - is the real asset at the early stage ㅤ 🔗 Resources Mentioned * bricks.sh [https://bricks.sh] - the guest's product * Retool [https://retool.com] - mentioned as a primary competitor and a tool Dario used extensively at his previous company * Lovable [https://lovable.dev] - mentioned as a vibe-coding alternative with no guardrails * Bolt [https://bolt.new] - mentioned alongside Lovable as a low-code/no-code alternative * Claude Code [https://claude.ai/code] - mentioned as an AI-native coding tool in the vibe-coding conversation * Supabase [https://supabase.com] - bricks.sh's current lead database integration * Stripe [https://stripe.com] - mentioned as a data source for future admin panel unification * Intercom [https://www.intercom.com] - mentioned as a data source for future admin panel unification * Linear [https://linear.app] - mentioned as a data source for future admin panel unification * Softup [https://www.softup.co] - Daniel's company, referenced in the conversation as a concrete example of the internal tooling problem