Lessons learned from studying 4k YC companies | Hacker News Reader

Lessons learned from studying 4k YC companies

amvizion.org

64 pointskovezd19 comments

I was expecting a GitHub repository with the data and processes. Instead, I am seeing this long text, (re)phrased by ChatGPT.

So here is a GPT summary of that long text

The text categorizes Y Combinator (YC) startups into three areas:

1. *Driving Efficiencies*: These startups improve existing markets with better, tech-enabled solutions, often disrupting current players.

2. *Removing Limitations*: These startups serve underserved communities or address new problems using existing technologies, such as FinTech in developing regions.

3. *Advancing Technology*: These startups push the boundaries of innovation with new technologies that transform industries, offering high rewards despite high risks.

The author critiques venture capital for being risk-averse and suggests a more proactive approach to nurturing deep tech and ambitious funding models.

1. Driving Efficiencies: Red Ocean - Sequioa Arc's "Hair on Fire" - Symbol.vc's "Consensus Markets"

2. Removing Limitations: Blue Ocean - Sequioa Arc's "Hard Fact of Life" - Symbol.vc's "Pre-Consensus Markets"

3. Advancing Technology: Sequioa Arc's "Future Vision" - Symbol.vc's "Non-Consensus Markets"

---

The Arc Product-Market Fit Framework

https://www.sequoiacap.com/article/pmf-framework/

Introducing Symbol: not seeking consensus

https://medium.com/symbol-vc/introducing-symbol-not-seeking-...

This is actually really helpful thank you

This was (partially) researched and written by ChatGPT, which probably explains why it contains no actual data or examples. The analysis is interesting, but entirely abstract.

Was it? That would make sense but I also don’t want to shit on an undergraduate for writing a very good undergrad paper.

"For the analysis, I used an LLM to embed the companies' value propositions in a vector space, ran a clustering algorithm on top of the embeddings, manually analyzed the results, and used ChatGPT to edit the draft."

This is the perfect use case for LLMs and Data Analysis in general.

I'd legitimately pay money for a similar productivity tool.

Companies born from HN comments should be their own category :)

I don't even know whether "a company's value proposition" in text actually represents its value proposition.

Tracks, so are most startups these days

> The most common mechanism for creating a venture-backed business is by bringing efficiency to existing markets. This approach is the simplest and least risky because the demand for the product already exists. The promise is to deliver a product that is quantitatively better than what currently exists. There is often little technological uncertainty, as existing technology is applied to a new domain (low R&D).

Great insight from the pile of data. I know a lot of fellow founders failed miserably to conclude on this same lessons.

This insight is more familiarly stated as:

take an existing budget line item

do not try and combine line items

do not cross operational boundaries in the customer

Self-fulfilling prophecy.

YC startups get pre-seed funding, ridiculously good deal on their seed, access to the alumni to market, free promotion on HN etc.

Which means they are going to have a 100x better chance of surviving until PMF versus someone who is bootstrapped or has limited access to capital.

So really what you're measuring isn't what makes a good startup but rather what type of startups get you into YC. And that has changed significantly pre and post Garry Tan taking over as CEO.

Now the statistics show you want to be based in SF, team of 2-3, 30 and under and building something involving LLMs. Which is kind of understandable given that we are in a gold rush period.

I’ve never understood the notion that YC is a good deal on the seed. Traditional series a VC, which I did, was far more capital for the same eventual percentage. You’re going to do a venture round. If 125k or whatever is significant to you, I guess it makes sense.

Series A today requires minimum $1m/year in ARR.

That's far beyond what YC companies are able to achieve in a few months.

So they are raising at a Seed level which YC companies get a good deal on because YC is effectively running their fundraising for them e.g. Demo Day, vetted investors, optimised process etc.

Series A does not require $1M/year. That's just for low-tier stuff.

Would love to see more of the output data. - How many win/loss clusters did the data produce? - In vector space what was the seperation between B2B and B2C companies? - Did you normalize against size of win/market-cap or anything else?

The point about not being able to create demand is absolutely vital to understand.

This analysis sounds very "smart" but in my opinion, contains little of substantial value.

Just make something a lot of people want, but can't currently get. That's the recipe.

Reply on news.ycombinator.com