Shrey Jain

YC’s best advice is “default alive” - have enough cash to survive without the next raise. Great advice… nobody f*cking does it. Paul Graham (the founder of YC) once said he’d raise $500K, get ramen profitable, and probably never raise again. The guy who’s seen more VC deployed than ANY human alive basically says he wouldn’t use it more than he absolutely has to. But the reason no one listens to this advice is because founders are delusional about their own companies. We think what we have is better than what it actually is, and we can’t help ourselves. Our startups are our babies. We have a totally unrealistic and out-of-touch view of them. It’s way too easy when you have a little bit of traction to go out there and raise way too much money. Then you’re set on a path to burn forever. “Default alive” is a great concept. Maybe 1/100 YC companies actually do this. YC knows it, and they fund them anyway.

140

50 Comments

A lot of VCs use founder age as a proxy for how well they’ll leverage LLMs and the latest AI. In my experience, the stronger signal is how technical they are. Technical founders understand how LLMs work and can exploit them fully; non-technical founders see magic, not mechanisms.

49

11 Comments

The difference between companies that are figuring out AI coding and those that aren’t is pretty stark. The ones that are figuring it out are concerned with: - dominating their industry - product velocity - development talent - building their own agentic workflows The ones that aren’t: - licenses (lol) - listening to a vocal minority (bc what if they leave!) - not stepping on any toes - traditional SDLC metric tracking

30

16 Comments

If you’re a first-time founder, there’s one ritual you can’t afford to ignore. Every day at 5 PM, Srinath and Matt at Regie.ai sit together for an hour with a cup of coffee and no agenda. It's space to surface whatever's bubbling up. They started this habit in the early days and kept to it. Databricks did weekly dinners for years with all five co-founders. Sometimes work-focused, but not always. Their goal was the same: create room for what isn't being said. Most founders talk constantly - in Slack, in standups, in product reviews. But those conversations are too transactional, too focused on output. You need unstructured time and the permission to say the uncomfortable thing. Room to address what's wrong before it becomes unsolvable. This is a powerful way to prevent a silent breakdown between leaders. As a startup grows, even the most amiable founders develop differences. With added responsibility and pressure, these differences turn into conflicts. A one-hour ritual can prevent that. My recommendation: Block the time, and keep it sacred. Founder fallout kills more startups than product missteps. If an hour a day - or a dinner a week - keeps that bond intact, why wouldn’t you make it non-negotiable?

392

22 Comments

If your studio builds deep-science ventures and your financial model assumes 12-month build cycles, you have the wrong model. Deep science ventures — biotech, advanced materials, hardware, climate tech — require 18-24 months of hands-on building before they're ready for external investment. The team needs lab access, regulatory navigation, and prototype iterations that don't compress into a quarter. This changes the studio economics completely. A SaaS-oriented studio can run 5-8 ventures simultaneously with lean teams. A deep-science studio is limited to 2-3 companies per lifecycle, each demanding significantly more capital and attention. The cost per venture is higher. The time to first revenue is longer. The follow-on capital requirements are larger. None of this means deep-science studios are bad businesses. It means they're different businesses that require different models: - Longer fund horizons (10-12 years, not 7-8) - Higher reserves per venture (2-3x a typical software build) - Institutional co-investment from day one — universities, national labs, federal grants - Stage gates tied to technical milestones, not customer acquisition metrics The studios that get this wrong usually start by importing a playbook from someone who builds SaaS companies. The pitch decks look great. The portfolio projections assume cadences that deep-science ventures can't hit. Then reality arrives around month 14, and the fund math doesn't work. Build the model around the science. Not the other way around.

14

1 Comment