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

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.

Founders scale faster on Deel. Set up payroll for any country in minutes, hire anyone anywhere, and get visas handled fast, so you stay focused on scaling. Deel takes care of onboarding, HR, IT, EOR, benefits, and compliance, so your team can grow without borders.

It’s why more than 40,000 fast-growing companies trust Deel to move fast.

Visit https://www.deel.com/dayone

Weekly Substack: https://georgiehealy.substack.com/

Linkedin: https://www.linkedin.com/in/georginahealy/

Instagram: https://www.instagram.com/georginahealytech/

In 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

Damian Naughton is the General Manager of ANZ at ElevenLabs, the voice AI company now past US$500 million in ARR, just four years after its 2022 founding. Formerly a regional VP at Slack and an early team member at Sydney's Hyper Anna (acquired by Alteryx), Damian is leading ElevenLabs' new Sydney office and tripling the local team as enterprise demand for voice AI accelerates.

In this episode, Damian and Georgie Healy cover why ElevenLabs is planting a flag in ANZ now and the local customers already on board, including Xero, Employment Hero, Heidi Health, Australia Post, and Andromeda Robotics' aged care companion robot Abby. Damian unpacks the company's high performance culture and the "11x yourself" philosophy, why intrinsic motivation beats top down pressure, and how he co-designs stretch goals with new hires. They get into voice cloning and the ethics behind it, how little audio you actually need to clone a voice, and how ElevenLabs handles consent, safety, and scammers, including a clever "reverse" use case that keeps fraudsters on the line to gather intel for banks. Damian also shares his view that enterprise AI needs to be led by the business rather than tech, why the BPO and consulting worlds are facing real disruption, and his take that tall poppy syndrome is a scourge holding Australian tech back.

Tune in for a candid conversation on voice AI, building a high performing team, and how enterprises should prepare for a near future where your own AI assistant takes action on your behalf.

Time Stamps

00:00 – Intro: Your AI Assistant Will Soon Take Action For You

02:17 – AI Hack of the Week: A Michael Caine Voice Butler On The Fridge

04:57 – ElevenLabs Launches in Australia and New Zealand

07:14 – Why Voice AI Is Ready Now (Xero, Employment Hero, Heidi, Andromeda)

10:52 – From Journalist to Voice AI: How AI Is Changing the News

13:40 – Inside Hyper Anna, the Aussie Startup Acquired by Alteryx

15:31 – ElevenLabs' High Performance Culture and the "11x Yourself" Rule

25:38 – The ElevenLabs Founding Story: Two Polish Founders and Bad Dubbing

28:42 – How Will Australia React to Voice AI?

31:24 – Voice Cloning Ethics, Consent, and Celebrity Voices

34:12 – How Little Audio You Need to Clone a Voice

37:17 – Voice AI and Scammers: How ElevenLabs Fights Fraud

42:05 – Can AI Bring Back Music From Artists Who Have Passed?

45:11 – Can Enterprise Companies Actually Adapt to AI?

53:05 – Advice for Heads of AI: Let Business Lead, Not Tech

56:39 – Rapid Fire: Tall Poppy Syndrome, Hiring, and Spicy Takes

1:02:05 – Where to Follow Damian and ElevenLabs

Founders scale faster on Deel. Set up payroll for any country in minutes, hire anyone anywhere, and get visas handled fast, so you stay focused on scaling. Deel takes care of onboarding, HR, IT, EOR, benefits, and compliance, so your team can grow without borders.

It’s why more than 40,000 fast-growing companies trust Deel to move fast.

Visit https://www.deel.com/dayone

Weekly Substack: https://georgiehealy.substack.com/

Linkedin: https://www.linkedin.com/in/georginahealy/

Instagram: https://www.instagram.com/georginahealytech/

In 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

Mentioned in this episode:

Deel x PX_Post Intro

Akshay Kothari is the COO and co-founder of Notion, the workspace platform now used by over a hundred million people. Before Notion, he co-founded Pulse, the newsreader app built as a Stanford class project that Steve Jobs name-checked on stage at WWDC 2010 before LinkedIn acquired it. He joined Notion in 2018 when the team was fewer than ten people, and in this conversation with Georgie Healy he traces that journey and where knowledge work is heading as agents take centre stage.

Akshay shares his AI hack of the week, turning a screenshot of restaurant recommendations into a shareable Notion database, and explains how the unit of work has shifted from taking notes to simply having a chat. He unpacks the design obsession behind Notion's identity, the block architecture that lets anyone build their own tools, and the new Developer Platform that brings outside agents like Claude and Codex onto Notion's context graph. He paints a picture of a "factory of agents" working round the clock while humans move to reviewing and applying taste, makes the case for model optionality and cost control, and shares his rule for custom agents: macro delegate, then micro steer.

Founders scale faster on Deel. Set up payroll for any country in minutes, hire anyone anywhere, and get visas handled fast, so you stay focused on scaling. Deel takes care of onboarding, HR, IT, EOR, benefits, and compliance, so your team can grow without borders.

It’s why more than 40,000 fast-growing companies trust Deel to move fast.

Visit https://www.deel.com/dayone

Weekly Substack: https://georgiehealy.substack.com/

Linkedin: https://www.linkedin.com/in/georginahealy/

Instagram: https://www.instagram.com/georginahealytech/

In 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

Mentioned in this episode:

Deel x PX_Post Intro

This is a #paidpartnership with Commonwealth Bank

Blair is the Chief Engineer of Generative AI at Commonwealth Bank, overseeing nine to ten teams building the AI platform that powers Australia’s largest bank and its millions of customers. Including me, a Dollar Mite since primary school.

His origin story is not what you would expect. He was a self-described hacker who grew up clicking through every system configuration setting he could find on his mum’s school computers after hours. That curiosity took him from building on GPT-2 before ChatGPT even existed, to the heart of one of Australia’s most important institutions.

In this episode we get into:

• The litmus test he uses to spot bad AI use immediately
• Why context matters more than prompting
• What an AI platform actually is and why your company probably needs one
• Why software engineers need to take more accountability for what they build
• The single security mistake most people are still making
• Why it is absolutely not too late to figure AI out for yourself

Practical, candid, and full of things you can action today.

00:00 Intro

03:05 What Does a Chief Engineer of AI at a Bank Actually Do

06:02 How CBA Collaborates Across Hundreds of Engineering Teams

08:56 How to Keep Up With AI Tools Without Getting Overwhelmed

12:04 Why Tool Consistency Matters in Enterprise AI

14:56 Why Blair Shares His AI Insights Publicly

17:51 What Being a Hacker Really Means in Tech

20:50 How Blair Hudson Went From Startup AI to Commonwealth Bank

23:59 Why CBA Starts Every Sprint With a Real Customer Call

27:06 Why Over-Engineering Is Killing Your AI Projects

29:47 Where AI Is Actually Moving the Needle in Banking Right Now

32:15 Why the Year of the Agent Was Overhyped

34:41 How to Build AI Fast Without Cutting Corners on Safety

39:28 What Is an AI Platform and Why Does Your Company Need One

41:49 How CBA Handles Shadow AI and Tool Adoption at Scale

45:09 Why Context Is More Important Than Prompting

50:29 The Number One Security Mistake You Are Still Making

52:30 The Easiest Way to Start Using AI to Save Time Today

54:05 Why You Should Never Ship AI Output Without Reviewing It

55:38 Why It Is Not Too Late to Learn AI

Founders scale faster on Deel. Set up payroll for any country in minutes, hire anyone anywhere, and get visas handled fast, so you stay focused on scaling. Deel takes care of onboarding, HR, IT, EOR, benefits, and compliance, so your team can grow without borders.

It’s why more than 40,000 fast-growing companies trust Deel to move fast.

Visit https://www.deel.com/dayone

Weekly Substack: https://georgiehealy.substack.com/

Linkedin: https://www.linkedin.com/in/georginahealy/

Instagram: https://www.instagram.com/georginahealytech/

In 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

Mentioned in this episode:

Deel x PX_Post Intro

"Can does not imply should."

That one line from Dr Simon Longstaff cuts to the heart of everything wrong with how the tech industry is currently building AI. The Executive Director of The Ethics Centre and one of Australia's most respected moral philosophers joins Georgie for a conversation that is equal parts grounding and mind-expanding, and one of the most important episodes the show has produced.

Simon's path to becoming Australia's foremost ethics expert is not what you would expect. He left school at 16, cleaned toilets on a remote island in the Gulf of Carpentaria, drove ambulances, became a paramedic and fire officer, and was eventually adopted by one of the clans of the Anindilyakwa people. It was at the end of a ship loading wharf that an indigenous elder taught him something about seeing patterns in the world, a lesson he has carried through 35 years of philosophical work and only recently realised had shaped everything.

In this episode he unpacks why ethics is not an optional extra bolted onto technology but the foundation it has to be built on, why the pharmaceutical approval model could be the blueprint for governing AI, and why "necessary fictions" mean that CEOs deploying AI are responsible for outcomes they literally cannot understand. He also makes the case that the coming wave of job displacement does not have to be a catastrophe, and explains what ancient Athens and pre-colonial Indigenous life have to do with universal basic income.

Founders scale faster on Deel. Set up payroll for any country in minutes, hire anyone anywhere, and get visas handled fast, so you stay focused on scaling. Deel takes care of onboarding, HR, IT, EOR, benefits, and compliance, so your team can grow without borders.

It’s why more than 40,000 fast-growing companies trust Deel to move fast.

Visit https://www.deel.com/dayone

Weekly Substack: https://georgiehealy.substack.com/

Linkedin: https://www.linkedin.com/in/georginahealy/

Instagram: https://www.instagram.com/georginahealytech/

In 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

Mentioned in this episode:

Deel x PX_Post Intro

When Katelyn Lesse was leading engineering at Stripe, she noticed everything around her shifting because of AI. So she left, and joined Anthropic. Today she leads platform engineering at one of the most important AI companies in the world, which is why we were so thrilled to host her for Anthropic's first ever interview in Sydney.

Katelyn does not lead with technical jargon. She leads with a question every builder needs to sit with right now. Are you building on the exponential, or are you stuck on the linear? Most people, she says, are already further behind than they realise.

In this episode she shares the exact framework she uses to think about building in the AI era, why your frustrations with Claude are actually signals you are onto something, and what it really means to be AGI pilled inside Anthropic.

00:00 Intro

02:53 AI Hack: Using Claude to Pick Wine

03:42 Katelyn's Background and Path to Anthropic

05:32 Why She Left Stripe to Join Anthropic

11:22 What Building on the Exponential Means

11:48 Why Australia Is a Top AI Market

13:05 Is the Claude Obsession Healthy?

15:20 The Framework for Building AI Products

16:38 Why Evals Are Non-Negotiable

18:18 Inside Anthropic's Developer Platform

20:11 What Is Harness Engineering?

25:13 AI Security, Sandboxes and Human in the Loop

29:09 Cloud Managed Agents Explained

31:56 Open Source MCP vs Closed Claude

33:48 How AI Is Changing Engineering Teams

36:24 How to Choose the Right AI Coding Tool

39:12 The End of the Traditional Tech Team

44:30 The AI Magic Moment Everyone Remembers

45:30 What It Means to Be AGI Pilled

47:21 Rapid Fire With Anthropic

49:56 Advice for Developers Building With AI

Founders scale faster on Deel. Set up payroll for any country in minutes, hire anyone anywhere, and get visas handled fast, so you stay focused on scaling. Deel takes care of onboarding, HR, IT, EOR, benefits, and compliance, so your team can grow without borders.

It’s why more than 40,000 fast-growing companies trust Deel to move fast.

Visit https://www.deel.com/dayone

Weekly Substack: https://georgiehealy.substack.com/

Linkedin: https://www.linkedin.com/in/georginahealy/

Instagram: https://www.instagram.com/georginahealytech/

In 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

Mentioned in this episode:

Deel x PX_Post Intro

Anna Zhou and Vivian Shen, the co-founders of Toastie, join Georgie Healy for one of the warmest and most personal conversations the show has had. Two software engineers at Google by day, they have quietly built one of the most thoughtful health tracking apps in the world by night, all without spending a single dollar on marketing.

Toastie was born from a problem they were both living. Anna was diagnosed with an autoimmune disease and handed a few photocopied sheets to figure out the rest of her life from. Viv has been managing PCOS for years, experimenting with medications and diets on her own. They realised the tools available simply were not built for people like them, so they built one themselves. Today Toastie helps users track symptoms, food, body signals, lab reports and scans, surfacing the patterns and irregularities that would otherwise go unnoticed.

In this episode they unpack why almost one in two Australians live with a chronic illness but no one talks about it, why slapping AI on everything is the wrong instinct and how they decide which features actually need it, and the cold LinkedIn email that landed them their first global partnership before they even had a product. They also share why ChatGPT and Gemini are not enough when the stakes are this high, what their users actually write to them in the feedback form, and the story of the user who quietly security audited their app and was so impressed they wrote in to tell them.

Plus the early hackathon they won by faking the backend in real time, why they call themselves boomers when it comes to social media, the worst startup advice they have ever received, and a special offer just for In The Blink of AI listeners.

🎁 Use promo code ITBOA2026 to get a 90 day free trial of Toastie

🍞 Find your Toastie personality: https://toastie.au/quiz

00:00 Intro

02:20 AI Hacks and Life at Google Sydney

07:00 How They Met and Their Hackathon Wins

10:51 What is Toastie and Why It Matters

13:50 The Personal Stories Behind the Product

16:23 How They Use AI (And Where They Don't)

17:37 The Cold Email That Landed a Global Partnership

20:00 Why General AI Models Aren't Enough

23:25 Building, Prioritising and the Competition

25:13 Why They Refuse to Call Themselves an AI Company

27:57 Trust, Security and User Feedback

30:00 Going Viral With Zero Marketing Spend

33:33 Handling AI Hate Online

35:43 Rapid Fire and What's Next

38:45 Special Offer for Listeners

Founders scale faster on Deel. Set up payroll for any country in minutes, hire anyone anywhere, and get visas handled fast, so you stay focused on scaling. Deel takes care of onboarding, HR, IT, EOR, benefits, and compliance, so your team can grow without borders.

It’s why more than 40,000 fast-growing companies trust Deel to move fast.

Visit https://www.deel.com/dayone

Weekly Substack: https://georgiehealy.substack.com/

Linkedin: https://www.linkedin.com/in/georginahealy/

Instagram: https://www.instagram.com/georginahealytech/

In 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

Mentioned in this episode:

Deel x PX_Post Intro