Key Takeaways

  • Adobe's blocked 20 billion dollar acquisition forced Figma CEO Dylan Field to keep building independently just as generative models began reshaping software.
  • OpenAI's overnight release solving over 700 famous mathematical problems showed Field that automated reasoning is quickly eliminating historic intellectual mysteries.
  • Software founders face an immediate fork: use machine intelligence to pursue difficult questions, or surrender active judgment to automated suggestions.
  • As automated coding lowers the cost of writing software, individual taste, visual design, and brand identity become the only remaining defenses against commoditization.

The Night OpenAI Solved 700 Math Problems

Dylan Field met Jason Calacanis in Napa under unusual circumstances. Regulators had just blocked Adobe from buying Figma for 20 billion dollars. Instead of dwelling on the regulatory collapse, Field spent the conversation processing a technical shock from the previous evening. OpenAI had dropped research papers resolving more than 700 famous mathematical problems at once.

Watching machines solve problems that had puzzled human minds for decades produced conflicting reactions. “Nothing hit home for me more than the math drop last night from OpenAI of 700 plus papers on all sorts of very famous problems,” Field said. “That was a show of something. I had a lot of emotions.”

Part of that reaction was grief. When an algorithm cracks a legendary problem overnight, the romance of human striving disappears. “Grappling with that mystery is something that is special,” Field remarked. “I'm both excited for the progress and like all the things it'll make possible, and also like I don't think it's wrong to mourn the loss of a little mystery.”

The Drift Toward an Algorithmic Hive Mind

The danger for modern operators is not that machines will refuse to help. The danger is that builders will stop thinking entirely. When answers arrive instantly, the friction required to develop original conviction vanishes. Teams end up accepting whatever average answer a prompt produces.

Field sees this choice confronting every creator. “As a human in general, pre-AI, post-AI, you have a choice. Do you want to consume or create? Do you want to take the easy path or do you want to climb the mountain? And what I'm seeing happen right now a lot is people are letting AI think for them.”

Surrendering your thinking creates software that feels like everything else. If every startup feeds identical questions into identical models, every product converges on the same consensus design. Field argues that real differentiation requires deliberate resistance against that pull. “My entire life has been about figuring out how to think,” Field explained, warning against builders casually “joining the hive mind.”

Taste and Willpower Beat Cheap Code

The rise of generative tooling has triggered endless warnings about an impending software crash. Field dismisses the extreme panic, but he acknowledges that technical implementation has lost its pricing power. When code generation costs drop toward zero, writing syntax no longer functions as a moat.

Competitive advantage shifts toward elements that machines cannot synthesize from training sets: personal taste, distinct design principles, and emotional brand resonance. Anyone can prompt a model to assemble a functional interface. Very few know which interface is worth shipping.

Retaining that edge comes down to human drive. “Everyone can be a maker. Everyone can go do bold things,” Field said. “It's about willpower and it's about curiosity. And if you're curious and you're willing to do the work, I believe.”

What to Do With This

Audit your team's output across the past two weeks. Look at the last five feature specs or architectural proposals submitted by your product team. Flag every paragraph or wireframe that was lifted directly from an LLM prompt without revision, and force the owner to rewrite it using an explicit, contrarian opinion on what the user actually needs.