Oct 26, 2013 · 29m · y-combinator
Office Hours at Startup School 2013 with Paul Graham and Sam Altman · Y Combinator
Paul Graham · 7m spoken Ryan Petersen · 4m spoken Nick Winter · 3m spoken Karen X. Cheng · 2m spoken George Saines · 2m spoken Finbarr Taylor · 2m spoken Sam Altman · 1m spoken
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
At Startup School 2013, Paul Graham and Sam Altman host live office hours, evaluating three early-stage startups: CodeCombat, GiveItOneHundred, and Flexport. Through interactive Q&A sessions, the founders receive real-time feedback on product mechanics, user acquisition, monetization, and long-term expansion strategies.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The partners as informed peer 5.6 Guest teaching 4.0 Guest disagreement 2.4 The partners pushing back 4.0
speaking balance: gold is the partners, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 24:08 Petersen rejects the Uber comparison
Petersen bluntly dismisses Graham's repeated comparison to Uber, correcting him that Flexport holds the actual brokerage license rather than merely dispatching third parties.
Hardest push from the partners ▶ 12:04 Graham dismantles waitlist gatingGraham refuses the premise that GiveItOneHundred needs more product polish before onboarding users, telling the founders that strong engagement proves they should open up immediately.
Biggest teaching moment ▶ 25:00 Proprietary database of all US importersPetersen educates the hosts on how public import filings and his prior data business give him an exhaustive database of every prospective US importer.
The partners hold their own ▶ 28:02 Graham and Altman check startup market framingAltman and Graham apply investor discipline against Petersen's 2.3 trillion dollar logistics figure, pointing out that overly massive TAM numbers alienate sophisticated investors.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The partners as informed peer | Guest teaching | Guest disagreement | The partners pushing back | Why |
|---|---|---|---|---|---|---|
| CodeCombat Pitch and Core Mechanics | 6 | 3 | 2 | 4 | Graham and Altman probe CodeCombat's core mechanics, questioning how gamification can force mastery of advanced programming rather than just memorizing library calls. Nick and George explain their progression mechanics and the immediate user engagement from their launch. | |
| Beta Feedback and Game Aesthetics | 5 | 4 | 2 | 3 | The interviewers explore beta testing learnings and infrastructure bottlenecks after a Reddit spike. George and Nick share how non-programmers struggle with basic syntax like semicolons, while the hosts offer thoughts on gaming industry practices. | |
| Advisor Feedback and Level Editor Tools | 7 | 3 | 2 | 4 | Graham gives advice on startup founder profiles and hiring while Altman challenges their recruitment-based monetization model by noting recruiters make unrealistic promises. The founders detail their level editor tooling and past startup background. | |
| Surprise Acceptance and GiveItOneHundred Pitch | 6 | 3 | 2 | 5 | Graham and Altman surprise CodeCombat with immediate batch acceptance before bringing on GiveItOneHundred. Graham pushes back on keeping a 4,300-person waitlist locked up, arguing that if engagement is already good, they should open up. | |
| Product Engagement and Video Constraints | 5 | 4 | 3 | 4 | Altman presses Karen and Finbarr on why they didn't test sharing on YouTube directly during the beta rather than hosting solely onsite. Karen clarifies their viral video strategy versus their daily web interface. | |
| User Experience and Narrative Potential | 6 | 3 | 3 | 5 | Graham dismisses the founder's idea that enterprise founders like Phil Libin would use the app to document startup struggles, redirecting them toward personal categories like baby milestones. Graham shares a personal anecdote about parenting memory decay. | |
| User Consistency and Community Support | 6 | 2 | 1 | 3 | Finbarr explains user retention data and community cheerleading features around difficult challenges. Graham advises them against prematurely optimizing for any single vertical before empirical usage patterns emerge. | |
| Modernizing Logistics with Online Dashboards | 5 | 6 | 2 | 3 | Ryan Petersen introduces Flexport, explaining customs brokerage regulations and workflow automation to an inquisitive Graham and Altman. Ryan details licensing requirements, FBI checks, and shipping volume economics. | |
| Regulatory Moats and Customer Acquisition Advantage | 4 | 7 | 4 | 4 | Ryan corrects Graham's analogy of being an Uber marketplace, clarifying Flexport is itself the licensed broker of record. Ryan also details his unique data asset of 300 million import manifests to acquire customers directly. | |
| Market Scaling, Sales Strategy, and Session Conclusion | 6 | 5 | 3 | 5 | Altman and Graham caution Ryan against citing trillions in TAM, advising focus on bottoms-up market reality. Altman challenges Ryan on building a direct outbound sales force to scale customer conversion. |
Statements from this episode (34)
Winter: CodeCombat is a multiplayer game for learning to code
“So we are building a multiplayer programming game for teaching people how to code.”
Winter: CodeCombat's launch traffic overwhelmed servers for 24 hours
“We can only serve a certain fraction of the traffic that we're getting, and that's been going on for 24 hours.”
Winter: CodeCombat Focuses on Core Programming Before Software Engineering
“So the software engineering part of learning to be a developer, that's something we can add later, focusing on core programming for now.”
Winter: Hard Game Levels Motivate Users to Learn Advanced Concepts
“And generally, if you make the levels hard enough, which you're able to do when they have a reason to complete it, they try really hard, then you can get them to do harder and harder stuff as a natural progression of the game.”
Paul Graham: Formal notation is the single biggest obstacle for beginner programmers
“The concept of formal notation is the single biggest obstacle.”
Saines: CodeCombat servers hit capacity at 15,000 visitors and 200 concurrents
“So we maxed out the server at 15,000 people, we had 200 concurrent, but we really don't know, because we were actually, people were just getting four or fours.”
Winter: CodeCombat plans to open-source within a few months
“Yeah, so we're thinking open sourcing in the next couple months to really capitalize on the interest.”
Paul Graham: Long corporate tenure is a negative signal on founder applications
“If we've, if we were looking at someone, someone's application, and they worked for a long time for a large company, That's actually bad to us, because like, the best startup founders probably could not stand that.”
Graham: CodeCombat is accepted into Y Combinator live during Startup School
“You didn't realize that, but that was your Y Combinator interview. You're in the next batch.”
Cheng: GiveItOneHundred has an invite-only beta of 50 people
“We have invite-only beta of 50 people”
Taylor: GiveItOneHundred has around 4,300 people on its waiting list
“Yeah, it's like 4300 or so on a waiting list.”
Cheng: GiveItOneHundred eliminated photo support because videos were more interesting
“We started off as a photo and video site, but then we cut out videos, we cut out photos because the videos were more interesting.”
Taylor: GiveItOneHundred gets 1,000 views and 20-30 daily returning users
“So the view counts, we're seeing about a thousand views a day, ah, and we have roughly between 20 and 30 of our kind of small group of users are coming back to the website every day.”
Cheng: GiveItOneHundred plans to turn user challenges into viral YouTube videos
“But we want to use YouTube, we want to piggyback off of YouTube as a marketing channel the same way we did with my video, so we'll take really compelling 100 day challenges, and we will turn it into a viral video, put it up on YouTube, and say, made with 100.”
Cheng: Initial dance challenge video achieved three million views
“Well, I guess our test for it is my video, which has three million views and was shared widely.”
Karen X. Cheng: GiveItOneHundred caps video clips at 10 seconds
“We cap it at 10 seconds, because I have a short attention span, and I'm building this for myself.”
Graham: Founders will not use GiveItOneHundred to document startup crises
“I don't think he would have used your system. No, seriously, not for, like, starting a startup.”
Graham: A child's current memories constantly overwrite parents' recent memories
“One of the big problems about being a parent is, The memories of the current kid overwrite the memories of the more recent, of the sort of recent past.”
Taylor: GiveItOneHundred users average around 18 video uploads each
“So out of the users that we have, there's an average of, ah, around 18 videos uploaded per user.”
Graham: Early-stage startups should stay empirical rather than over-optimize too early
“Don't be, don't, like over optimize too early. Like, let it grow into whatever it's gonna grow into. Maybe it'll end up being kids are the big thing, or maybe not, who knows. But be empirical about it, and don't, like, wire in some outcome too early.”
Petersen: Flexport is first licensed US customs brokerage with modern web app
“We're the first licensed U.S. Customs brokerage built around a modern web application.”
Petersen: Obtaining customs brokerage license took Flexport years
“It took us years to get the license.”
Petersen: Importing goods can require filing up to 120 different forms
“Based on what the product is, there could be, like, a 120 different forms you have to file, so we have to take the, what the product is, determine which forms are needed, fill those out for the, ah, customer, file them electronically with the government.”
Petersen: Flexport employs licensed customs brokers rather than removing humans
“We're not removing the human elements, so we're pairing you with a licensed customs worker. We have customs workers on staff, and it's more of a communications tool to, like, enable it.”
Petersen: U.S. customs brokers charge $100-$300 across 30M annual shipments
“Usually it's not dependent on the value of the goods, but it's like between a 103 hundred dollars per shipment, and there's thirty million shipments that enter the U.S. Every year and are filed with a customs customs entry.”
Petersen: Flexport has a waiting list of about 300 companies
“Yeah, we're trying, we actually have a waiting list of like, 300.”
Petersen: The world's largest oil company joined Flexport's waitlist
“The biggest oil company in the world signed up.”
Petersen: US customs rules blocked centralized software clearance before 2007
“And actually until recently it wasn't possible to clear shipment except at your local port. So if you built a software startup to do this, you could only help people importing into the port of Oakland, unless you had an office in every port. In 2007.”
Petersen: Prior startup aggregated 300M U.S. shipping manifests for subscription sales
“Every time you import a product into the U.S., it's public record, that product, and my last company actually sells that data. We aggregate every time you import something. We've collected three hundred million of those shipping manifests and sell subscription…”
Petersen: Flexport expects roughly 75% gross margins per clearance
“The gross margin should be about 75%.”