Offcuts

On this episode, Offcuts went live again, this time in London, on stage at Forza Wine in Soho in front of a room of commerce and brand leaders.

I'm joined by three people who think about taste for a living: Eleanor Warnock, journalist and advisor to Bek Ventures, who started her career at the Wall Street Journal; Max Ottignon, co-founder of Ragged Edge, the branding agency behind Wise, Monzo and Granola and Natalie Varma, co-founder and CPO of Parfumerie, who previously led product at Farfetch, Net-a-Porter and Condé Nast.

The question for the morning: when every brand is using the same models, what's left to differentiate on?
We split it into three parts.

Voice: who writes your brand, and where Eleanor lets AI in and where she keeps it out.
Brand: why Max thinks the winners of the AI race will be decided by brand, and why Granola's identity was built as the opposite of tech slop.
Curation: what happens to discovery when AI does the choosing, and how Natalie is tackling a category where you can't smell the internet.

We also get into whether AI will ever have taste, why the hardest part of brand building is deciding what to leave out, and why a little more friction might be good for us.

Recorded live with our sponsors Harper, Nolo and Centra.

Enjoy the show.

On this episode I sit down with Alex O'Byrne, an independent writer, builder and investor focused on AI-first companies, and my former co-founder at the Shopify agency we built and sold together. Alex now runs Alpha Factory, an AI investment company, from Miami. This is his fourth time on Offcuts.

This time we go deep on what happened after he built an AI hedge fund, Money Monster, that read earnings transcripts to try to predict stock moves, and took real losses when the Iran war broke its macro signals. He walks me through how he rebuilt it as Alpha Factory, multi-strategy, multi-instrument, and increasingly run without him, after realising his own risk calls were often the weak link in the system.

From there we get into the organisation of the future: why Alex thinks the one-person billion-dollar company is mostly rage bait, what he thinks is actually achievable solo, and the role he calls the orchestrator, the generalist who directs a stack of AI agents and supplies the taste the machines still can't fake.

We also get into why he thinks frontier labs' safety messaging is more political positioning than genuine alarm, and what a dog park in Miami taught him about how obsolete the standard thirty-minute Zoom call is starting to feel.

Enjoy the show.

On this episode I sit down with Alister Hewitt, Technology and Operations Director at Rat & Boa and co-founder of Prebuy.

Alister has been with the brand since the very beginning. He built the first website for a friend launching a clothing label, and within a couple of months the orders were flying in.
He describes what Rat & Boa actually is: born on Instagram in 2014, back when the platform had no product tags and no stories and brands used it as a mood board, built around the founders' aesthetic, worn organically by the likes of Bella Hadid and Alexa Chung, and still run by a far smaller team than people assume.

Most of the conversation is about the problem that comes with being a viral brand. Collections sell out in a couple of days, the rebuy window runs six to eight weeks and sometimes ten, and back-in-stock alerts were only converting at five to ten per cent. Alister's view is that preorders look like a product page feature and are really a supply chain one. Change the CTA and you just push the problem into the warehouse, the merch team and the customer service inbox. That is where Prebuy came from.

The second half is about how far AI has gone inside the business. A dev team that is almost AI native, with Cursor as the harness. A CRO audit skill the digital team built themselves. Gorgias and Shopify MCPs in customer service. A server in the building running local models, because the token bill is climbing. And why Alister thinks agentic shopping is sitting exactly where mobile commerce sat a decade ago.

Enjoy the show.

On this episode I sit down with Eleanor Warnock, a writer, journalist and communications expert based in London who has worked for The Wall Street Journal, the Financial Times's Sifted, and the Reid Hoffman-backed AI company Every. She's currently an advisor to Bek Ventures, ranked the top-performing VC firm globally since 2010 by HEC-Dow Jones, and holds an MBA from INSEAD.

Eleanor is about to launch her own venture, and this conversation ranges across the state of journalism's business model, the myth of the one-person billion-dollar AI company, and where AI and taste collide.

We get into why she thinks journalism is heading toward a handful of consolidated giants and a long tail of one-person outlets, why Every's hybrid of media and consulting works, and why even she needed a lawyer to check the incorporation documents Claude wrote for her. We also dig into taste: why she thinks it's really about cultural status, why the tech world wants it and can't build it, and where in her own daily workflow she lets Claude in and where she keeps it firmly out.

Enjoy the show.

On this episode I sit down with Olivier Legris, co-founder of Alter, a native macOS AI assistant that refuses to bet on any single model, and founder of Clauders, the global Claude Ambassador community.

Olivier's thesis is simple: the model doesn't matter to most people, the task does. Alter started as a frustration with Siri and has become a Mac-native, model-agnostic assistant with its own router, Best, that picks the right model for the job. When Alter shipped that router, Olivier expected a fight over which model was best. Instead, barely 10% of users ever touch the model picker. They just want the task done.

We get into why Mac is quietly the best platform for building AI products, the challenge of building on a market that moves faster than any roadmap can keep up with, and Olivier's read on the "Linux of AI" moment happening with agent frameworks right now, powerful, but still too complicated for anyone outside the bleeding edge.

We also talk about the Clauders community, how it grew from 50 ambassadors to a genuine cross-border network, and the 90/9/1 rule that keeps it alive. And we close on the big picture: open source versus proprietary models, on-device AI, and whether Anthropic and OpenAI will actually IPO this year.

Enjoy the show.