Depth Over Breadth
Earlier this week, a friend who leads startup partnerships at one of the major labs asked me whether the founder journey is easier or harder than it was 5-10 years ago.
I started Petal ten years ago and Modern Life five, I’ve run my n of 2 experiment within the window. My answer is both, but depth is still all that really matters.
It’s harder than ever to get attention and plow through the noise to actually sell your product. Every category has a dozen funded (and un-funded) competitors and every buyer is drowning in AI-generated outreach. Reputation, trust, and distribution matter more than ever.
It’s easier than ever to start a company, harder than ever to stand out. Ten years ago, before we raised a dollar at Petal, my cofounders and I spent nine months building relationships with card networks, issuing banks, and processors. Investors asked hard questions and we couldn’t fake the answers, we had to be steeped in the problem. The game was ask the right questions, and grind to find the answers. Now it’s mostly just asking the right questions given AI can give a decent answer to most questions. It’s an opportunity to grind on the answers that aren’t yet trained into the models.
Building is the same story. I used to spend nights and weekends grinding code, visualizing the whole system in my head. Now when I get a pitch, sometimes I’ll go to Claude Code and have it take a stab at architecting the same exact thing, and sometimes I’ll have it implement a POC. Most things are easy to build now.
Money is more available than ever at the early stage, which means more companies get funded, which means more lower-quality businesses get venture dollars than before. This journey is already brutal when you have a great team, product, and GTM, it’s way harder when you’re up against five companies that got funded without proving anything.
And it’s not just other startups. It used to be incumbents vs. startups. Now it’s a free-for-all: incumbents vs. startups vs. forward-deployed shops. Anthropic and OpenAI are hiring forward deployed engineers and competing head to head with the startups building on their platforms. Consulting was always competition in a sense, but it was never this prevalent.
Every founder has AI, which means the old pitch-level signals are broken. It used to be that depth showed, you could tell within twenty minutes whether someone had lived in a problem. Now it’s genuinely hard to distinguish a founder who’s spent two years in a space from one who had Claude prep them the night before. The questions that used to separate them don’t anymore. Same thing with fundraising: we used to spend weeks and weeks iterating on the deck and honing the story, these days a significant portion of companies have a memo at best, which isn’t the same work as building a deck. It’s like Eisenhower said, “plans are worthless, but planning is everything.”
If what you’re building has shallow domain specificity, you’re in trouble because anyone can build it, including me, in an afternoon between meetings, and an FDE team will happily build it custom for your customer. But if you’re encoding real depth in a specific area, the stuff you can’t prompt your way into, you have a much better chance. Software was never really the moat, anything was always buildable, but now that it’s trivial to build, it’s important to have deep domain knowledge encoded into a product no one else knows to build.
Depth over breadth.
The Alchemist Discovering Phosphorus, Joseph Wright of Derby, 1771.
