
"Shit at the speed of light is still shit."
That one line from Pip captures the entire philosophy behind Springboards, the AI company she co-founded with Amy and Kieran that is quietly pushing back against what the rest of the industry is doing. The three of them join Georgie Healy for one of the most thought-provoking conversations the show has had about what AI is quietly doing to creativity, and what it takes to build a model that breaks the mould.
Pip and Amy never planned to start an AI company. Both worked in advertising and got laid off within three weeks of each other, which led them to accidentally build the first version of Springboards themselves to solve a problem they kept running into: existing AI tools were not helping them do creative work better, they were making everyone's creative work look the same. Kieran joined as their technical co-founder and together they have now released Flint, a divergence model designed to break the AI hive mind.
In this episode they unpack why 69 out of 70 language models will tell you that time is a river, why mainstream AI has converged into one gray mush of sameness, and why the scariest part of this might be that most people will not even notice. They also get into how they built Flint to score 7.5 on novelty bench when the frontier models score ones and twos, why the smallest possible model was always the goal, and why they deliberately avoid making the tool feel too polished.
Plus why humans are evolutionarily lazy and what that means for our brains in the AI era, the unexpected analogy about sourdough and alcohol that changes how you think about creativity, and the honest reflection from all three founders on being the self-loathing AI company in a space full of hype.
Chapters
00:00 โ Intro
02:22 โ Introducing Flint and the convergence problem in AI models
04:50 โ Why Springboards is uniquely positioned to solve creative AI
07:30 โ What entropy actually means in language models
09:44 โ Real examples: random cars, pizza toppings, and where to holiday
12:36 โ Why this matters for the advertising industry (and everyone else)
15:29 โ Inside Flint: how to fine-tune a model for divergence
18:38 โ Doubling the score on Novelty Bench (and what that even means)
23:35 โ Try Flint yourself: who it's for and how to access it
26:10 โ Cognitive atrophy, taste, and keeping humans in the creative loop
35:04 โ Choosing a tech provider as an early-stage AI startup
38:20 โ What actually matters for founders in the sasspocalypse era
41:12 โ Rapid fire: copyright, cover shoots, Eumundi markets, and self-loathing AI
More Episodes You Might Like
Let's talk
Turn podcasting into pipeline
We help founders, funds and operators build trust, authority and deal flow with a show tailored to their market.
Win better deals and stay topโofโmind with founders.
Close more deals and build a category you own.
Reach founders and operators with a show they trust.















