The "levered beta" insight is spot on - when tech improves this fast, traditional moats crumble.
But there's a key difference between being early and having real distribution.
Harvey looks like ChatGPT in a legal costume, but they seem to be doing something much smarter: creating workflow dependencies. Once firms build their document templates, train staff, and structure approval processes around Harvey, switching becomes painful regardless of model improvements.
The real play isn't alpha vs beta. it's managing the transition. Ride the LLM wave while building infrastructure that matters post-commoditization. Use current users to map future workflow needs.
Replit's infrastructure bet suddenly looks genius. When AI-generated code becomes table stakes, advantage goes to whoever controls deployment and scaling.
This "levered beta" moment won't last. Folks that are gaining edge are using today's hype to embed themselves so deep into workflows that removal becomes unthinkable.
Most people still think we’re in 2010 when it comes to startup valuation. In 2025, it’s no longer just about the product or even current MRR. The business models with AI are about perceived market dominance, founder leverage, and velocity, like signing football stars more for potential and audience pull than stats alone.
Valuation today reflects distribution, narrative, and user traction, not technical sophistication. MRR is becoming a vanity metric unless it signals network effects or market capture.
Just look at Windsurf after Devin bought it, it ranked higher than many more mature tools because of developer traction and social proof. That traction matters more than a polished architecture.
You can vibe-code a Facebook clone in a weekend, sure. But without users, without mindshare, without momentum, you’re not building Facebook. You’re building a ghost town.
The "levered beta" insight is spot on - when tech improves this fast, traditional moats crumble.
But there's a key difference between being early and having real distribution.
Harvey looks like ChatGPT in a legal costume, but they seem to be doing something much smarter: creating workflow dependencies. Once firms build their document templates, train staff, and structure approval processes around Harvey, switching becomes painful regardless of model improvements.
The real play isn't alpha vs beta. it's managing the transition. Ride the LLM wave while building infrastructure that matters post-commoditization. Use current users to map future workflow needs.
Replit's infrastructure bet suddenly looks genius. When AI-generated code becomes table stakes, advantage goes to whoever controls deployment and scaling.
This "levered beta" moment won't last. Folks that are gaining edge are using today's hype to embed themselves so deep into workflows that removal becomes unthinkable.
"when the market goes up 10%, your 2x levered beta portfolio goes up 20%. when it goes down 10%, well, that’s a problem for future you"
great line
Spot on - and very well written!
Most people still think we’re in 2010 when it comes to startup valuation. In 2025, it’s no longer just about the product or even current MRR. The business models with AI are about perceived market dominance, founder leverage, and velocity, like signing football stars more for potential and audience pull than stats alone.
Valuation today reflects distribution, narrative, and user traction, not technical sophistication. MRR is becoming a vanity metric unless it signals network effects or market capture.
Just look at Windsurf after Devin bought it, it ranked higher than many more mature tools because of developer traction and social proof. That traction matters more than a polished architecture.
You can vibe-code a Facebook clone in a weekend, sure. But without users, without mindshare, without momentum, you’re not building Facebook. You’re building a ghost town.
Interesting article but is comparing icon&cluely to lovable a good comparison? Claude was already good enough when lovable came out