Lessons from Our Y Combinator Batch

Y Combinator teaches you a lot of things. Most of it is advice you've heard before, delivered with enough conviction and context that it finally clicks. Some of it is advice you've never heard, and it changes how you work permanently.

Here's what actually stuck from our S24 batch, filtered through building DenchClaw.

The Advice That Actually Changed How We Work#

"Write code or talk to users. Everything else is a distraction."#

We heard this multiple times from multiple YC partners in different forms. The first few weeks of the batch, we were doing a lot of things that weren't those two. Planning documents. Competitive analysis. Pitch deck iterations. YC disabused us of this quickly.

The discipline they teach: every single day, you should either be shipping something or learning something from a user. If you end the day having done neither, the day was mostly wasted.

We've internalized this. We ship something every day. The gstack workflow we use — Think → Plan → Build → Review → Test → Ship — is designed around daily shipping as the default, not the exception.

The "do things that don't scale" insight is more profound than it sounds#

Paul Graham's essay is famous enough that most founders treat it as received wisdom without really understanding it. YC made it concrete.

The point isn't that you should do unscalable things instead of building. It's that the unscalable things are how you learn what to build. When we manually helped DenchClaw users set up their workspaces — literally on calls, walking them through the configuration — we learned ten things we could never have learned from analytics. We learned exactly what was confusing, what they were trying to do, what they expected the product to do that it didn't.

Those calls made the product better. The calls themselves don't scale. What they produced does.

Pivot early, not late#

We renamed and partially repositioned DenchClaw from Ironclaw during YC. The original framing was too narrow. Garry Tan's tweet and the Show HN response told us something important: people were excited about the broader "AI workspace on your Mac" framing, not just the CRM piece.

YC culture moves fast on pivots. The partners encourage you to update your mental model weekly based on what you're learning. If your users are consistently using your product for something different than what you built it for, that's signal — not something to fight.

What YC Got Wrong (Or Didn't Apply to Us)#

The "growth at all costs" pressure#

YC is, at its core, a machine for producing high-growth venture-backed startups. That's fine. But the pressure toward user growth metrics can conflict with building something people deeply trust.

For DenchClaw, trust is the product. People are giving a local AI agent access to their most sensitive data — their contacts, their deals, their notes, their browser. Growth that comes at the cost of trust would destroy what we're building.

We've been more deliberate about growth than the typical YC trajectory suggests. We build in public, ship consistently, and let compounding organic discovery do a lot of the work. It's slower. The foundation is more solid.

"Don't spend time on open source, spend time on the product"#

We heard versions of this. We disagreed. Our open-source model is inseparable from the product — it's how users trust us with their local data. An open-source AI agent running locally is categorically different from a closed-source cloud service, in the eyes of users who care about privacy.

YC's advice makes sense for most companies. It didn't apply to us.

The Things Nobody Tells You#

How much the network compounds#

Three months after batch ends, you realize the lasting value of YC isn't the funding — it's the 100+ other founders you now know who are working on adjacent things. The information flow between YC alumni is genuinely valuable and can't be replicated.

Demo Day pressure is useful#

Demo Day forces clarity. You have two minutes to explain your product to investors who know nothing about it. That forcing function makes you articulate your value proposition more clearly than any amount of internal strategy documents. Do it early, do it often, and share the feedback with your whole team.

The best advice comes from other founders, not partners#

The YC partners are smart and well-intentioned, but they're working from pattern matching on thousands of companies. The most useful advice we got came from other S24 founders who were working on adjacent problems and had just figured something out that applied to us.

Build relationships with other founders in your batch. Not for networking — for thinking together.

How DenchClaw Came Out of YC Different#

We shipped more in the three months of the batch than in the six months before it. The constraint of weekly group office hours — where you have to show progress — forces a cadence.

We also came out with a clearer thesis: the local-first AI workspace for founders and knowledge workers, not just a CRM. That's a bigger bet. The YC environment gave us the confidence to make it.

Most importantly: we came out with users who genuinely care about what we're building. That's harder to manufacture than anything else. Those users shape the product more than any strategic document.

Frequently Asked Questions#

Was YC worth it for an open-source, local-first product?#

Yes, decisively. The network, the forcing function for shipping, and the credibility with early adopters all mattered. The funding matters less for a capital-light product, but the community matters enormously.

What's the most underrated thing YC does?#

Forces you to talk to your users every week and report back on what you learned. The accountability structure alone changes your habits.

How did YC affect your relationship with your co-founder?#

It creates pressure that reveals what kind of team you are. We came through it stronger because we'd already established trust before applying. My advice: don't apply to YC if co-founder dynamics are shaky. The pressure will make cracks visible quickly.

What would you do differently going into YC?#

Start the user conversations earlier. We were doing a lot of building in the first few weeks before we'd really understood what users needed. In retrospect, the user conversations should have been week one, not week four.

Ready to try DenchClaw? Install in one command: npx denchclaw. Full setup guide →