S26 Day One: What the First Week Actually Decides

Miles: The S26 batch just started. Decisions went out June 5th, roughly 1% of 10,000 plus applicants. And right now, those founders are sitting in San Francisco staring at a blank calendar trying to figure out what to do before office hours even exist. Grant: The gap between getting accepted and understanding what you're actually supposed to be doing, that's the part nobody talks about. Miles: Right, so today we're dropping into that exact moment. We've got a founder walking us through the first hours, not the pitch. It's not the narrative they'll tell at Demo Day, the unedited version. Grant: And what I want to pull at because I've seen this pattern is how much got decided before any YC structure was there to guide them, who to trust in the room, which signals to act on that all get set early. Miles: Yeah, yeah, and we're also going to get into the RFS, the Summer 2026 Request for Startups, which the VC corner covered and is a different animal than what we've seen before. 15 categories. Eight of them require hardware or capital. Grant: For what's supposed to be a software accelerator, that's a real statement. Miles: I've seen this play out a hundred times. Founders read the RFS as a checklist. The ones who read it as a conviction map are the ones who show up to office hours with something to say. Grant: The math on that distinction tends to be brutal and pretty fast. Miles: And we close on a founder who's still inside S26 right now, sitting on a call that hasn't happened yet. And yet, build the chip or build the scheduler, no answer exists. Batch just started. Grant: So, code open, found a voice, let's go. Miles: You know, what nobody told me is that the room is quieter than you'd expect. Grant: Yeah. Miles: You've got 200 founders roughly all sitting there. And I remember thinking, I've flown here, I've quit my job, I've told everyone I know that this is the thing and the partner just starts talking. No countdown, no fanfare. And I'm looking around at all these people thinking, everyone in here has a version of that story. Grant: Did it feel like validation or pressure? Miles: Both, like simultaneously. You want it to feel like an arrival, and it does a little, but the second they start talking about what the batch actually demands, you realize the acceptance letter was the easy part. Grant: That tracks. Miles: I talked to one S26 founder who described it as, she said it felt like the first day of a job where everyone else already seems to know what they're doing and you're the only one who missed the orientation. Grant: Yeah. I mean, that psychological thing is real. You spend months just trying to get in. The application, the interview. Capwave published data saying YC accepts somewhere between 1.5 and 2% of applicants, lower acceptance rate than Harvard. Miles: Right, so you've cleared that, and then you walk in and the room just continues. The batch starts. YC's site is pretty plain about it. Three-day in-person kickoff, you meet your group partners, you meet your cohort, that's it. That's the structure. Grant: And the gap between what you imagine and what that actually looks like. Like, I think that's the thing that catches people. Miles: Because the fantasy version has some kind of ceremony to it, like someone hands you the playbook. But the VCcorner site's piece on the Summer 2026 RFS, they make this point that the batch is more technically ambitious this cycle, more specific about where YC thinks the next generation of companies is actually coming from. So you're not just starting, you're starting inside a room full of people who were selected because YC is placing a directional bet. that right now. Grant: And most of those founders don't fully know what that bet is yet, day one. Miles: Day one. You're in that room, the partner's talking, and you've got maybe three months to Demo Day. Founders I've spoken to, they describe the 11 weeks leading up to Demo Day as the most productive stretch of their lives, and I believe that. But Hour One, Hour One is disorienting. Grant: You either figure out how to use that disorientation. or it uses you. Miles: So what does figuring it out actually look like when you're inside it, before office hours even start, before you've had a single conversation with your partner? What gets decided in those first 72 hours, and who's making those calls? So the first real decision, I want to be specific here, not the one founders rehearsed for their YC interview, the one that walked in uninvited on day two. Grant: And this is where it gets murky fast, because when you talk to these S26 founders now, everything sounds intentional. It sounds like a plan. Miles: Right. Hindsight is a hell of an editor. Grant: So where were they actually working with? Miles: Almost nothing. I mean, they've been in the room maybe 48 hours. Office hours haven't started, they don't have a partner relationship yet, they're surrounded by what? A couple hundred other companies, all of whom look like they already know exactly what they're doing. Grant: That number matters. Young Combinator's S26 batch, 250 to 300 companies per published numbers. You're one of them. And from inside that room, the other 249 all seem further along than you. Miles: I've talked to founders across multiple batches who describe that exact exact feeling. Everyone's nodding like they have traction. Nobody's admitting they have no idea what week one actually requires. Grant: So what did they decide? Miles: The two founders I spoke to, different companies, different categories, both made essentially the same call. They picked who to talk to before they had any framework for who mattered. Grant: That sounds small. Miles: It's not. Think about what information costs you in that environment. You've got maybe a hundred A hundred conversations you could have in week one, the ones you choose shape what you think is true about your market for the next 12 weeks. Grant: And the founders knew that at the time? Miles: No, that's the thing I kept pushing on. I'd ask what data did you actually have when you made that call, and the answer was usually some version of a hunch, a conversation from the week before the batch, maybe a Slack message. Grant: So they were filling in missing information with whatever was closest. Miles: With whoever was in the room—which is a very specific sample. Grant: That's the port I want to sit on, because what they're describing—deciding who to trust, who to ignore, which signal to act on-that's not a day two conversation. That's the batch. That's twelve weeks of compounding off of first instinct. Miles: And what's wild is neither of them flagged it as a decision at the time-one of them literally said, I wasn't deciding anything, I was just eating lunch with someone. Grant: Yeah, that's how the expensive ones always start. Miles: The YC program is structured. Groups of six to ten companies, a dedicated general partner, weekly office hours. But that structure doesn't kick in immediately. There's a gap. Grant: And what happens in the gap? Miles: That's exactly what the partner sees that the founders don't. One vantage point on a hundred companies making the same call without knowing they're making it. So what a partner sees from week one and what a founder sees, those are two completely different movies. Grant: That's the thing, right? Each section is six to ten companies. Your partner is watching all of them simultaneously. You're watching only yourself. Miles: According to YC Combinator's own program structure. Each general partner hand selects a small group every batch and meets with them weekly, so the information asymmetry is built in. The partner has the comparison set; you don't. Grant: And the pattern they are watching for in week one isn't momentum, is whether founders can tell the difference between being busy and actually moving. Miles: I've seen this play out a hundred times in other contexts: someone fills the calendar, fires off emails, takes every intro call. All three weeks in, nothing has changed. Grant: Confusing motion for progress. That's the one. And the thing is, it feels identical from the inside. Miles: So how does a partner catch it early? Grant: The question they're apparently asking at office hours, YC's own framing, is basically, what's actually holding you back from moving faster? Not what did you do this week? What is the obstacle? Miles: Which is a very different question. Grant: Completely different. And it separates the founders who know the... Know their own bottleneck from the ones who are just narrating their schedule. Miles: Narrating their schedule. Yeah. Grant: Now flip it to the RFS side. The VC Corner published a breakdown of all 15 S26 categories, AI-Native Service Companies, Company Brain, Inference Chips for Agent Workflows, Counter-Swarm Defense, Electronics and Space. It spans AI, hardware, Defense, agriculture, space. Miles: 15 categories and the partners who wrote it are now that are now in the room with S26 founders, they can see in real time who's building towards those gaps and who drifted in with something adjacent. Grant: Adjacent is a polite word. Miles: It is. And YC Combinator's own RFS page is explicit. This isn't a trend report. These are gaps that are now solvable. The framing is operational, not aspirational. Grant: Which means a partner in week one already has a mental model. Is this founder building towards a real gap? Or are they building towards what felt exciting in their own heads six months ago? Miles: Narrating their schedule, yeah. That gap between the RFS signal and what's actually being built, that's probably the most useful data the partner has before any office hours conversation even starts. Grant: Right, and it raises the obvious question: what does the RFS actually tell you if you read it the way a partner reads it-not as a check list, but as a set of conditions? Miles: Which is exactly where we're going, because you've been treating it like term sheet language, and it changes what you see in it. So the VCCorner piece on the S26 RFS, I pulled it apart like a term sheet. 15 categories, 8 of them require capital, hardware, or actual atoms. That's Grant: Wow. Miles: the biggest physical world emphasis in RFS history. Grant: And most founders walking into Week One probably skimmed it. Miles: Right, they skimmed it, but the partners didn't write it to be skimmed. Every category has a named partner attached. Garry Tan on agriculture robotics. X, Diana Hu shows up three times, inference chips, semiconductor supply chains, AIOS for companies. That's not an accident, that's a bet. Grant: Three entries from one person is a statement. Miles: It's a conviction map, and when a partner sits down with you in week one, they already know whose name is on the category closest to what you're building. Okay, but, and I push back on this a little, most founders in S26 didn't get in because they perfectly matched a category. to category. They got in because they had something real. So where does the RFS actually land for them in week one? Grant: That's the tension. Reading it as a checklist is the wrong move. Am I AI native services? Am I company brain? That's not the question. Miles: What's the question? Grant: Well, the question is, what problem does Diana Hu actually lose sleep over? Because company brain as a category is YC saying out loud that every company's most critical knowledge. How to improve a special pricing escalation How to handle a refund escalation It's scattered across Slack histories and the heads of people who might quit tomorrow That's the gap The category name is just the label Always Miles: So you're reading underneath the label. Grant: I've done this in deal making for 20 years The document tells you what they want to fund The conversation tells you what they're afraid won't get built Miles: And if you're a founder who doesn't map cleanly to any of the 15? Grant: YC is explicit about this. They fund far more companies outside the RFS than inside it. But the founders who understand the thesis behind the categories, not just the category names, those are the ones who walk into office hours and sound like they've been thinking at the same altitude as the partner. Miles: There's a version of this where a founder in S26 right now is sitting on something genuinely in the inference chips for agent workflows lane. I mean, current GPUs hitting 30-40% utilization on agentic workloads. That's the gap Diana who is describing, and they don't know it yet because they haven't read closely enough. Grant: Or they know it and are not saying it clearly. Miles: Which brings us to one founder who's in that exact position right now. Grant: Yeah, and they haven't decided yet. Miles: That decision is live. We'll get into it. So this founder, the one in the inference chips lane, they've got a real fork in front of them right now. Grant: Yeah, walk me through it. Miles: Okay, so according to the VC corner's breakdown of the S26 RFS, current GPUs are hitting 30 to 40% of peak utilization on agentic workloads. The work is bursty, context switching is brutal, and Diana Hu is explicitly asking for chips designed around the agent. Agent loop. Fast context switching. KV caches that persist across an entire execution graph. This founder is sitting right in the middle of that gap. Grant: So what's the fork? Miles: Do you build the chip or do you build the systems that makes existing chips more efficient? Two very different companies. Grant: Oh, that's a brutal one. That's like, that's a two-year difference in timeline. Miles: Minimum. And the batch is, what, barely a weekend. Grant: Yeah, no pressure. Miles: And I think, look, if I'm watching this from the outside, the thing I'd be clocking over the next two weeks is whether their first technical hire is a chip architect or a systems engineer. That's the tell. Grant: Mmm, because one of those hires closes the door. Miles: Completely. You hire the chip architect, you're committing to a three- to five-year capital story. You hire the systems engineer, you're betting you can extract enough performance from existing silicon that customers will pay for it now. Grant: From a number standpoint, the way I look at it, can you get a design partner to pay for a proof of concept in the next 60 days? If the answer is no, you probably don't have the chip play yet. You have a thesis. Miles: That's exactly the line. A thesis isn't a company. Grant: YC will pressure test that fast. Week two office hours are not soft. Miles: The thing is, NVIDIA paid $20 billion for Groq because they saw this exact gap coming. That's not a small signal, but seeing the gap and being the company that fills it are two different problems. Grant: Yeah. Miles: This founder knows the architecture; they know the agent execution model; the v The VC Corner called it "the founder who understands both chip architecture and how agents actually execute has a rare window right now. Grant: Key word being window. Miles: So that's where it sits. The decision hasn't been made. S26 is three months long and it started this week. We don't know which door they walk through. Grant: We genuinely don't. Miles: And neither do they. All right, that's the episode. Grant: Yeah-and I keep coming back to that room: no countdown, no fanfare-the partner just starts talking. Miles: That detail stuck with me too. You spend months getting in and then day one it's just... work. Grant: And somewhere in that room right now, there's a founder who still hasn't made the call: chip or scheduler. The batch just started. Miles: That's a cliffhanger I don't want to let go of. Grant: Same. Look, if you know a YC founder in their first year who'll be willing to tell that story honestly, send them our way: yearone at heymeadow dot com. Miles: And if this episode helped you think differently about what actually happens inside that room, leave us a review, it helps. Grant: Miles: good one today. Miles: Yeah, more soon. Thanks for listening, Everybody.