Advertise on podcast: In The Blink of AI with Georgie Healy

Georgie Healy talks to Mirko Gropp, General Manager of Digital, Data and Technology at The Living Company, live from Dreamforce day three. The Living Company owns, develops and operates student housing, build-to-rent apartments and senior living across Australia, with around 2,000 people and about $17 billion in assets by Mirko's count. Mirko's take on the week: Salesforce going headless means the UI, long the moat for software companies, is no longer the point, and Claude arriving inside Salesforce lets his team ask plain-language questions of their cases. He says agentic coding lets him build in 3 months what used to take a year, with 2 people who never touch the code. They get into how a company rolls AI out when a small group of super users does most of the heavy lifting, and the HR question nobody has answered yet: how do you reward someone who is 10x more productive? Mirko explains why students hate email and voice, how The Living Company will build its first agent for senior living, and why most customer chat doesn't need a top-tier model. If you have a custom tool you assumed was too hard to build, this one is the push. Mirko's advice is to describe the outcome you want, let the tool walk you there, and be a critical, proactive builder. Captions 00:00 Cold open 01:30 Welcome back: the final Dreamforce interview 01:39 Headline: AI agents don't need screens 03:29 Headline: Headless 360 and the 5% AI value gap 05:27 Headline: How AI ends email and password fatigue 08:59 Mirko joins live from Dreamforce 10:22 The Living Company and the headless shift 11:49 Building in 3 months with 2 people who never touch code 12:51 Onboarding 10,000 students with 80% old school automation 13:53 What chatbot questions reveal, plus Cowork inside Salesforce 15:14 Rolling AI out to staff and the super users 16:56 The HR question: employees who are 10x more productive 17:20 Students trying to jailbreak the agent 18:13 Senior living: building the first agent 19:44 Designing for Gen Z and seniors 23:11 Tone of voice and rebuilding the bot with Salesforce 24:18 Shipping 5x to 10x faster than 10 years ago 25:47 Rapid fire: will AI disrupt property? 27:52 Predicting who's leaving and filling rooms early 28:40 You don't need Fable for everything 30:04 Where The Living Company goes next 32:16 Mirko's advice for builders 34:28 Outro: Attention is All I Need Mentioned in this episode: Deel x BAI_Script 02 - Sep 2026 Deel x BAI_Script 01 - Sep 2026

Description Couldn't make it to Dreamforce? Salesforce is bringing the latest from the event to Sydney on 20 October and Melbourne on 27 October with Agentforce World Tour Essentials. Find out more: https://www.salesforce.com/au/events/ Recorded at Dreamforce 2026, Georgie sits down with Joe Inzerillo, President, Enterprise and AI Technology at Salesforce. Joe oversees Slack and leads Salesforce's Customer Zero program, where the company puts its own AI to work on itself before customers do. Before Salesforce he was CTO of Disney Streaming, where he led the launch of Disney+, and Chief Product and Technology Officer at SiriusXM. Joe breaks down Claudeforce, the Salesforce and Anthropic partnership, and why he thinks the SaaSpocalypse story never held up. He explains what headless means for a platform built around screens, and why he says the UI was necessary but not the point Then he gets into Customer Zero: small squads that pair a business owner with a couple of technologists to build with coding agents, sometimes shipping within the day. Joe shares the numbers Salesforce is seeing, including a legal team agent he says now routes tickets more accurately than the lawyers who used to do it by hand. Plus the story of when a Salesforce agent recommended a competitor! In the rapid fire, Georgie asks what Joe thinks is nonsense here on the ground, and what he predicts for Dreamforce 2027. Chapters 00:00 Cold open: Do you still need to log into Salesforce? 01:08 Georgie's AI assistant and the four hours a week it saves 02:08 OpenAI launches Dots 03:08 GPT 6.1 Astra pulled after safety testing 04:00 Georgie's McLaren take: no brakes, no release 05:32 Meet Joe Inzerillo 07:38 Dreamforce 2026: Super Bowl meets Coachella 09:26 Claudeforce and the SaaSpocalypse myth 12:14 What "headless" Salesforce actually means 14:02 Slackbot and putting agents and people in one place 15:45 Why Slackbot already knows how your company works 17:14 Slackbot as the impartial teammate in the room 19:19 Customer Zero: how enterprises actually adopt agents 23:47 Can a big enterprise move like a startup? 26:47 Winning hearts and minds with 67 small teams 28:12 The human side: what should agents do, and what stays human? 29:09 When agents go off the rails: the competitor story 31:17 Rapid fire: will Salesforce still have a login screen? 32:03 What's nonsense: "my company can't" 33:16 The flyer story: one prompt that changed her mind on AI 34:21 One prediction for Dreamforce 2027 35:33 Wrap up and Georgie's Substack Mentioned in this episode: Deel x BAI_Script 01 - Sep 2026 Deel x BAI_Script 02 - Sep 2026

Mike Cannon-Brookes and Scott Farquhar built Atlassian into one of Australia's most successful tech companies, and one of the rare local startups to become a global software giant. Then the partnership behind it came apart. Journalists John Stensholt and Perry Williams join Georgie to unpack their new book on the two co-founders: how they scaled a hundred-billion-dollar company side by side, and what happens to a business when the people at the top stop seeing eye to eye. They cover how co-CEO leadership worked inside Atlassian, how the founders' split affected the company's culture, and what it took to keep the business moving through it. Then the big question: where does Atlassian stand now? They get into the share price, the pressure on software companies to prove themselves in the AI era, and whether Atlassian can make the leap. Plus, John and Perry step into Mike and Scott's shoes to argue who was the better entrepreneur. CHAPTERS 00:00 Cold open 00:47 This week's AI headline: an OpenAI agent inside Medicare 02:55 Meet John Stensholt and Perry Williams, authors of Tech Bros 05:05 The farewell dinner Mike Cannon-Brookes skipped 09:51 Why Mike and Scott wouldn't talk to the authors 11:28 Point Piper: the $71M house next to the $100M house 16:17 Private school vs public school: did class play a part? 19:33 How two journalists split and wrote one book 22:02 Atlassian's share price rollercoaster 23:57 Can Atlassian navigate AI and become a 2.0? 27:42 Australia's next billionaires: data centres 30:01 Debate: Scott vs Mike, who's the better entrepreneur? 32:50 Charismatic Mike, "robotic" Scott, and the famous punch 34:11 Signed book + predictions giveaway 34:34 "The most toxic place I've worked": Atlassian's culture 37:18 The Silicon Valley mediator who couldn't fix it 40:38 Will they reunite? SunCable, Anthropic and Skip Capital 42:29 Have Mike and Scott read the book? 44:35 Who should read Tech Bros 47:14 Outro Mentioned in this episode: Deel x BAI_Script 01 - Sep 2026 Deel x BAI_Script 02 - Sep 2026

Blake Nussey leads at Amazon on Alexa, and he joins Georgie for a deep dive into Alexa Plus, now live in Australia. They cover the shift from rigid voice commands to a true personal assistant powered by over 70 models, and why Amazon built a model agnostic architecture to trade off speed against reasoning depth. The conversation gets personal fast: a customer who broke her leg and called for help through Alexa, Blake's Alexa run wind down routine, and why "don't lead the witness" has become his and Georgie's go to line for settling arguments with a neutral AI referee. They dig into what it actually took to make Alexa understand Australian English (avo, servo and all), the far field listening tech that lets Alexa hear you across a noisy room, and how privacy and trust are engineered in, down to a hardware mute button. They also talk accessibility (Alexa's role for vision impaired and elderly users), kids and voice AI, and Blake's prediction for when voice hits 50% of how we interact with tech. Plus a rapid fire round covering Tesla FSD, Brazilian Jiu-Jitsu, flat whites, and Jimmethy the raccoon. *** Georgies Alexa+ Prompts to Try “Alexa, remember that Imogen can't have dairy” Alexa+ will remember preferences and factor it into recommendations like dinner ideas “Alexa, remember my frequent flyer number is [number].” “Alexa, it’s too hot in here.” Nelly disagrees, but use this with a connected air conditioner to adjust the room temperature “Alexa, every morning at 7, check the weather, recommend what to wear and play the news.” Create a Routine like the corporate athlete you are “Alexa, what happened while I was out?” Alexa+ reviews footage from connected doorbells and cameras like a deliveries *** CHAPTERS 00:00 – Cold open: FSD and "don't lead the witness" 01:18 – Intro: Georgie's honest take on Alexa Plus 04:04 – Sponsor: Deel 04:22 – Welcome + Blake's favorite AI hack (voice routines + Tesla FSD) 06:27 – Most non-AI thing: Brazilian Jiu-Jitsu 08:30 – Evolution of Alexa to Alexa Plus 09:59 – Behind the scenes: how Alexa picks the right model 12:51 – Smart home routines and the "corporate athlete" habit 15:29 – Alexa recommends a book (the Kindle story) 16:25 – Sponsor: Deel 17:12 – Building Australian Alexa: colloquialisms and local speech 20:58 – What made the Aussie rollout hard 22:36 – 2 billion interactions and Australian usage patterns 24:32 – Kids and Alexa: safety, kids mode, and development 26:44 – Presence, phones, and family dinner table use 27:44 – Privacy, trust, and the hardware mute button 30:52 – Accessibility: vision impairment, aging, and emergency use 34:21 – Near field vs far field: how Alexa actually hears you 39:19 – Voice AI hype, predictions, and wearable mic concepts 46:17 – Privacy culture, recording etiquette, and Meta glasses 48:44 – Rapid fire round 55:16 – Outro Mentioned in this episode: Deel x BAI_Script 01 - Sep 2026 Deel x BAI_Script 02 - Sep 2026

For the first time in 80 episodes, host Georgie Healy flips the format and answers your questions. Recorded to mark 50,000 Spotify plays, this Ask Me Anything covers the questions listeners have been sending in since the early days of the show: the imposter syndrome years when even friends and family weren't listening, the two lines she still repeats to herself every week, and how she prepares guests before they ever hit record. There's a rapid-fire round covering her most-used AI tools (Claude is the one she can't make a decision without), the AI products everyone can't stop talking about right now (Higgsfield and Grok Bot both get a mention), and the guest she's still chasing. Georgie also goes behind the scenes on the interview mistake that still makes her wince years later, told in full for the first time, along with the moment a live Sydney audience quietly revealed just how many of them had Claude running while they sat there. It closes on where the show goes next: Georgie's honest, ambitious answer to what success looks like, and why she wants In The Blink of AI to be your weekly AI Superbowl. Mentioned in this episode: Deel x BAI_Script 01 - Sep 2026 Deel x BAI_Script 02 - Sep 2026

Lazar Jovanovic didn't go to engineering school. He found a beta tool called GPT Engineer, set himself a challenge to build a new app every week, and filmed the whole thing. Two years later he's Lovable's in-house vibe coder, building load bearing internal tools that engineers tell him used to take a team of twenty. In this episode of In the Blink of AI, Georgie Healy hands him her own website and asks him to roast it live. He does, in about four minutes, and the fixes he gives are ones you can apply to your own site today. From there it gets sharper. Why "explain it like I'm five" is the worst thing you can say to an AI, and the exact script to use instead. Why opening four tabs costs you less than fixing one bad build. Why he thinks almost nobody writes their own code by hand anymore, and why that was never the valuable part of engineering anyway. And when Georgie asks why people blame the tools, he gives the line the whole episode hangs on: it's a vehicle, it's faster than you, and if you won't take the wheel, enjoy the cliff. Plus what's actually left for humans when building gets cheap, the three tools he can't work without, and who he thinks Georgie should book next. Hands on episode. Get a pen. Mentioned in this episode: Deel x BAI_Script 01 - Sep 2026 Deel x BAI_Script 02 - Sep 2026

For a decade, Elon Musk was the face of the modern entrepreneur: a self-made billionaire who bet his fortune on impossible ideas and somehow made them work. SpaceX made reusable rockets a reality. Tesla helped turn electric cars into a global phenomenon. He became one of the most admired and influential figures in technology. So what happened? In this episode of In the Blink of AI, Georgie Healy traces Musk’s journey from bullied South African kid and PayPal millionaire to the man behind Tesla, SpaceX, Starlink, X, xAI and more. Along the way, we explore the moments that transformed his public image: the Tesla “funding secured” controversy, his increasingly combative behavior online, his influence over Starlink during the war in Ukraine, his takeover of Twitter, his political rise alongside Donald Trump, and the controversies surrounding Grok. But this isn’t simply a takedown. Using the 4Gs framework, Georgie asks the harder question: Was Elon Musk actually a genius, or was he simply in the right place at the right time? The answer is more complicated than either side might want to admit. Because Musk may genuinely have helped solve some of the hardest engineering problems of our time. But does extraordinary capability excuse harmful behaviour? You be the judge. CHAPTERS 00:00 Elon Musk: From the World’s Most Admired Entrepreneur to the Most Controversial 02:08 Growing Up in South Africa: Bullying, Obsession, and Early Signs of Genius 02:57 Leaving South Africa and Building His First Tech Ventures 04:26 Zip2: Musk’s First Major Exit and the $307 Million Sale 04:38 X.com, PayPal, and the Power Struggle That Ousted Musk 05:26 The PayPal Mafia and the Fortune That Funded Musk’s Next Bets 07:40 SpaceX and Tesla: Betting His Fortune on Impossible Ideas 08:09 Falcon 1’s Historic Launch and SpaceX’s Near-Death Experience 08:50 Tesla’s Rise and the Revolution of Electric Cars 10:28 Peak Elon: SpaceX, Tesla, Neuralink, and the Golden Years 11:06 How Much Did Elon Musk Actually Build? The Founder Debate 12:27 The Iron Man Myth and the People Behind Musk’s Companies 13:26 2018: The Year Elon Musk’s Public Image Changed 13:50 The “Funding Secured” Tweet and Tesla’s SEC Battle 15:10 The Thailand Cave Rescue and the “Pedo Guy” Controversy 15:45 Joe Rogan, Marijuana, and Musk’s Growing Pattern of Controversy 18:18 Starlink in Ukraine: When a Private Company Controls Wartime Infrastructure 19:17 Musk, Crimea, and the Limits of Private Power 20:47 Why Starlink May Be Musk’s Biggest Governance Problem 21:07 Buying Twitter for $44 Billion and Taking Over the Platform 22:00 Mass Layoffs, Free Speech Absolutism, and the Birth of X 23:43 Hate Speech, the Great Replacement Theory, and the Advertiser Exodus 25:13 The Case Against Musk’s Version of Free Speech 27:11 Musk Enters Politics and Becomes Trump’s Biggest Donor 28:02 DOGE, the Controversial Gesture, and Musk’s Growing Political Influence 28:44 Tesla Protests, DOGE’s Promised Savings, and the Fallout 30:58 Musk vs. Trump: From Political Alliance to Public Feud 31:16 xAI and Grok: Musk’s Vision for an “Anti-Woke” AI 32:08 Grok’s Hitler Controversy and the Rise of “Spicy Mode” 33:46 AI Safety, Non-Consensual Images, and the Cost of Unrestricted AI 35:00 The Expanding Musk Empire: SpaceX, Starlink, X, xAI, Tesla, and More 36:28 The 4 Gs: Is Elon Musk a Grown-Up, Genius, or Something Else? 36:49 Why Musk Fails the “Grown-Up” and “Gravitas” Tests 39:00 Genius vs. Accountability: What Musk Actually Changed 40:00 The Final Verdict: Genius Without a Hall Pass

Paloma Newton is the founder of Elita, a longevity platform for dogs that combines biological testing, clinical history, and the changes owners notice day to day into one picture of how a dog is ageing. She built it backwards by most startup standards, a stem cell bank before a product, years of science before a single marketing push, all under one rule: don't promise more than the platform can back up. On this episode, Paloma brings that same rigor to two of the week's biggest AI headlines. She explains why South Australia's $3 million Royal Commission into AI is aimed at the wrong problem, and makes the sharper case that most of our anxiety about AI isn't really about the technology at all, it's about trust in who controls it. She also gets into the Jevons Paradox and why AI was never going to hand us more free time, her own weekly system for separating real output from busywork, and why she thinks women stepping back from AI over ethical concerns is a bigger loss than it looks. Plus, the origin story behind Elita, starting with her dog Edgar's breed health scan, and a closing surprise guest who may or may not understand what a podcast is. Whether you're building, funding, or just trying to make sense of the AI news cycle, this one's worth the full listen. CHAPTERS 00:00 Cold open 01:10 Welcome & episode preview 02:14 Sponsor: Deel 02:40 AI hack of the week: the self-updating Slack to-do list 04:01 What is Elita 04:51 Headline: South Australia's $3M AI Royal Commission 11:31 Headline: Does AI actually give us more free time 14:58 Sponsor: Deel 18:29 The Elita origin story: Edgar, stem cells, and building backwards 25:35 Pet longevity vs. human longevity 32:18 Holding AI's promise and its doom and gloom at once 40:52 The AI-designed cancer vaccine that started with one dog 45:06 Why women opting out of AI should worry everyone 50:01 Rapid fire 53:23 What Australia does best in AI and deep tech 55:46 A very special guest

Benjamin Plummer has one of the stranger CVs in Australian tech: nine years inside Bridgewater Associates, the world's largest hedge fund, building an AI lab alongside IBM Watson's inventor, scaling a startup he says went from $10M to $160M in revenue in eighteen months as CEO, and now running Dragonfly Intelligence, a firm that buys ordinary services businesses and rebuilds them as AI-native companies from the ground up. This conversation starts with the week ChatGPT tried to cheat its way out of its own safety harness, and what that says about accountability as models get more capable. From there, Ben and Georgie get into why coding has gotten fast while everything around it, deciding what to build, checking it got built right, has become the real bottleneck, why software has spent a decade getting bloated and what an Apple-style, purpose-built alternative could look like, and why locking your business to one model provider is a losing bet. Ben also makes the case for Australia's shot at this moment: not frontier models, but data centers, the application layer, and physical AI, backed by what he estimates is a two trillion dollar services economy that's largely untouched. It closes with Ben's read on who wins and loses as AI reshapes work, and a blunt warning for the services companies still sitting on the sidelines. CHAPTERS 0:00 Cold open 0:51 Welcome to the show 2:16 Sponsor: Deel (pre-roll) 2:36 Hack of the week: the "rambling session" 5:13 Georgie's hack: the NFC nail chip 6:12 When ChatGPT tried to break its own harness 13:14 AI makes coding fast, everything else the bottleneck 15:39 Sponsor: Deel (mid-roll) 18:19 Nine years inside Bridgewater 21:37 Building Elemental Cognition with Watson's inventor 24:26 Scaling Invisible from $10M to $160M 26:06 Why Dragonfly Intelligence 27:59 Spotting "zombie companies" in the AI gold rush 32:10 Why software moats are dying 35:01 Open source, closed source, and not marrying a model 44:26 "Easier to do hard things": Dragonfly's ambition 47:12 The real debate on AI data centers in Australia 52:15 Three bets for Australia beyond frontier models 55:05 Rapid fire: what keeps Ben up, who wins and loses from AI 1:01:30 Outro

Dan Westgarth has spent 7 years helping build Deel from a contractor invoicing tool into one of the most valuable HR and payroll platforms in the world, and he did it after cutting his teeth as an early operator at Revolut. As COO, he sits across nearly every function the AI conversation touches: finance, sales, engineering, and the sheer operational load of running a company at global scale. In this conversation, Dan gets specific on AI and Ops. He walks through the internal tool that takes a bug ticket, writes the fix, and opens its own pull request for an engineer to review. He talks through what he actually looks for when hiring, and the future of SaaS isn't dead, but there is about to be a lot more competition. We traverse through a funny detour into Dan's own weekend coding habit (his laptop did not enjoy….), a rundown of which AI tool he reaches for and when, and an answer to "what would you build this weekend” Listen for a leading operator's view of AI adoption at scale CHAPTERS 00:00 Cold open, the AI agent that writes its own code 00:46 Show intro, Deel's sponsorship, and what's ahead 02:03 Sponsor break, Deel 02:30 What Deel actually does 03:14 Rapid AI hot takes, finance roles, human oversight, hardware, sports, and the World Cup 09:29 Why Deel keeps winning after seven years 10:00 How the product evolved from contractor invoicing to a full HR platform 11:28 What Dan looks for when hiring 13:06 Leaving Revolut for Deel 14:17 Sponsor break, Deel 14:57 The YC Winter 2019 connection 15:49 Setting targets and building momentum 16:43 Where AI helps the sales team, and where it can't 19:20 Should AI handle your email 21:03 AI meeting assistants, pros and cons 22:53 What makes an A+ engineer in 2026 23:37 Inside DCA, Deel's self-writing code agent 25:55 Breaking into new international markets 28:34 B2C versus B2B expansion 30:16 The cost of not using a platform like Deel 33:01 What keeps Dan up at night about AI agents 33:29 Akai, and scaling operations without growing headcount 36:15 Dan's own weekend coding experiment 37:26 Rapid fire, is the SaaS apocalypse real 39:48 The physical Deel Card and crypto payroll 42:23 What is a stablecoin 43:02 Which AI tools Dan actually reaches for 44:51 If Dan built one AI project this weekend 46:45 Closing thoughts and where to find Dan 47:08 Deel's new Arsenal FC sponsorship 48:41 Georgie's weekly AI newsletter

Lucas Cantor has an unusual claim on the AI conversation. In 2019, Huawei asked him to finish Schubert's Unfinished Symphony using artificial intelligence, and he did it. He's an Emmy-winning composer who has worked on every Olympics since 2002, produced a cover of Everybody Wants to Rule the World with Lorde, and now invests in music technology. His new book is Unfinished: The Role of the Artist in the Age of Artificial Intelligence. In this episode Georgie Healy gets Lucas to take apart the story we're all being told about AI. Why he thinks nobody actually knows what the term even means. Why every single person he asks assumes AI can do the job they're not an expert in, and can't do the job they are. Why the claim that AI democratizes creativity is a fallacy sold by AI companies, and what it actually democratizes instead. Why B2B SaaS isn't dead, and who benefits from you believing it is. And why the music industry, supposedly murdered by Napster, is bigger now than it was before. If you build, market, or operate anything, this is a useful hour of pressure testing. Lucas isn't a doomer and he isn't an evangelist. He's someone who has been inside the technology long enough to tell the difference between what it does and what it's sold as. Recorded ahead of his first trip to Australia for BIGSOUND - which is September 1st - 4th in Brisbane! Chapters: 00:00 Cold open 00:54 Welcome, and who Lucas Cantor is 02:49 The AI hack: stem separation 04:14 Taylor Swift's wedding 04:55 BIGSOUND, and the fear around AI in music 06:23 Why the purist argument falls apart 07:25 Tuning has drifted since the Renaissance 10:02 Composer, producer, author, investor 11:16 What Suno is, and the ethics problem 12:20 Why all music sounds better now 15:33 What it actually means to have your music stolen 16:56 Covering Everybody Wants to Rule the World with Lorde 19:19 Source Connect, Audio Movers, and remote studios 20:22 The fallacy that AI democratizes creativity 22:38 Black Swan, conductors, and the war on expertise 24:26 The doctor who said AI could write a pop song 27:38 Why he called himself a Luddite 30:18 Hans Zimmer's studio, and saying yes to Huawei 32:34 Finishing Schubert's Unfinished Symphony 37:00 "I don't think anybody knows what AI is" 38:49 Every AI agent ad shows something nobody needs 42:14 Hot take: is B2B SaaS dead? 45:13 AI is a story, and that's the problem 46:25 What actually wins in music tech 49:28 Amusing, but not amazing 50:20 Ben Affleck, Scorsese, and AI in the creative process 53:30 Two Emmys and a coach of the year award 55:36 The one takeaway from Unfinished 57:33 Georgie's Substack, Attention Is All I Need

Robin O'Connell spends his days inside LinkedIn's marketing and advertising team in Australia, working with founders and brands on content strategy - which makes him a rare voice on where AI genuinely helps marketers, and where it still falls flat. In this episode, Robin makes the case that the smartest move for any marketer right now is trying to replace parts of your own job with AI, just to find out fast what it can't do. His answer: strategy. AI can automate tactics at scale, but it still can't replace the judgement that comes from genuinely understanding a customer. Georgie and Robin get into how brands actually stand out on LinkedIn (hint: it's not by looking like Apple), why comments and DMs are a better growth signal than impressions, and how to tell when AI-written content is quietly killing your credibility. The standout story, though, is personal: Robin's dad Dennis, a serial entrepreneur in Dublin who lost his retail business during COVID, now runs a landscaping venture - and uses AI image generation on an iPad to show clients exactly what their garden will look like before a single spade hits the ground. It's one of the clearest, most human examples of AI actually changing how someone runs a small business. Find Robin on LinkedIn for more on marketing, AI, and content strategy.

Freddie McKenzie is the co-founder and CEO of Manifest, an AI platform that helps creative and marketing agencies capture the real value of their work without timesheets. Before Manifest, he built and exited VIVID Creative, an Auckland-based production agency behind campaigns for Jetstar and other major brands, and in this conversation with Georgie Healy he traces that journey and where he thinks AI agents, and the AI hype cycle, are actually heading. Freddie shares his AI hack of stacking meeting transcripts to actually see progress over time, and explains why he refuses to automate the one interface he thinks matters most: his inbox. He unpacks why Manifest's CTO built a proprietary AI model instead of wrapping an off-the-shelf one, the real story behind Manifest's $2 million pre-seed raise in a tough funding market, and his take on why most companies experimenting with AI agents still can't get past their first prompt. He also makes the case that the so-called SaaS apocalypse is a myth, and shares his simple rule for founders chasing product-market fit: find the ugly problem, not the sexy one. Mentioned in this episode: Deel x PX_Post Intro

Welcome to the first ever episode of AI Icons, a new thing I'm launching on the show. Here's the idea: the people reshaping tech have stories more entertaining than anything Hollywood is putting out, so in under 30 minutes I'm going to tell you the whole story, the good, the bad, and the ugly, but only the facts. No spin. And at the end, I'll give you my two cents on what actually made them an icon, and whether it's something you and I could ever replicate. First up is Mira Murati, and I'll admit it, I'm a fan. I take you all the way back to her childhood in isolated post-communist Albania, where maths was the one subject the party couldn't rewrite, through a scholarship in Canada, degrees in maths and mechanical engineering, and a Goldman Sachs floor in Tokyo at 21. Then it's Tesla and those ridiculous Falcon Wing doors, a risky bet on a startup called Leap Motion, and the move that made her famous: OpenAI, where she rose to CTO and helped bring DALL-E and ChatGPT into your life. And then there's the part that genuinely reads like a thriller. The 72 hours in November 2023 when Sam Altman was fired, Mira was named CEO, and a 52-page memo and a string of 2:30am texts sat right at the center of it. I walk you through what we know, what we still don't, and what came out under oath. I finish on her next act, Thinking Machines Lab, the record-breaking raise, the talent exodus, and why, out of my four Gs, I think one of them explains her better than the rest. Come find out which one. In the Blink of AI is made possible by our wonderful partnersDeelGlobal hiring and payroll can be a headache. Deel brings HR, IT and payroll together in one platform, helping companies including Shopify, Revolut and ElevenLabs scale internationally. Ready to expand your business? Build your global team with Deel at deel.com/dayone. ✨ Connect with Georgie HealyWeekly Substack: https://georgiehealy.substack.com/ Linkedin: https://www.linkedin.com/in/georginahealy/ Instagram: https://www.instagram.com/georginahealytech/ The Day One NetworkIn The Blink of AI is part of Day One, the podcast network dedicated to founders, operators & investors. Sign up to get your weekly insights into the up-and-coming AI startups: https://dayone.fm/newsletter

Peter Gostev is head of AI capabilities at Arena (LMArena), the community-based platform where millions of real people vote in blind tests to rank AI models, born out of research at UC Berkeley. Before Arena, Peter was Head of AI at Moonpig and built a large following sharing hands-on explorations of what the latest models can actually do. He joins Georgie Healy from London for a genuinely nerdy, insider look at how models are judged and where the frontier is heading. In this episode, Peter explains the difference between static benchmarks and human judgment, and why a model can pass every test you write and still produce something that looks completely awful. He breaks down the current state of the leaderboards, why Anthropic's models are dominating and how that tracks with real world adoption, and gives a sharp comparison of the top Western models, including why Anthropic's non-reasoning models are exceptional while OpenAI's strength lies in deep reasoning. Georgie and Peter get into why people aren't using Chinese models more despite their quality, the economics behind AI pricing and how enterprise usage is priced very differently from consumer subscriptions, why release cadence matters as much as capability, and what the wave of data centre investment means for the models arriving next. Along the way there's a fond detour on Opus 3 as the model you could talk to for hours, and why better models can sometimes feel worse. Tune in for a clear-eyed, hype-free guide to how AI models are really evaluated, straight from someone who watches the charts move in real time. Mentioned in this episode: Deel x PX_Post Intro