The Network Effects Map continues to help @NFX founders see what others do not and max their efforts to build the dominant companies in their sectors.

@aref_vc What do you think founders care about the most from VC's?

Well done @MaxPGreenwald and @warmlyai . Another 3X, another milestone. Very fun to watch @NFX

C'mon people, are we really surprised by @deepseek_ai? Did you really not see it coming? From our @nfx blog in 2022: "The [core AI models] are going to be open-source and easy to use. Models is where the generative AI revolution started – but it’s not where it will end. It’s not going to take long, maybe 3 years, for models to approximate each other and reach human limits. The ability for these models to differentiate themselves stops at those limits. Data network effects asymptote over time. An AI model that is 5% or even 20% percent better than the competition is a pretty slim defensibility. The AI models will be switched out at will. Which of course ... commodify them.” There will be several more 100X reductions in speed and cost in the next two years, and eventually it will run on CPU's and we won't need GPU's, and it will run on your smartphone and laptop without heating it up. Note that @Meta pulled the same trick as DeepSeek but not as well (yet) with Llama which is to say "Fuck, I lost, so I'll blow the whole market up by doing opensource 6 months after you. If I can't win, I'm going to keep you from winning." Many players have it in their strategic insterest to open source AI capability. Meta is one. China is a second. There are many more. It's inevitable. Hasn't this been obvious from the beginning? The defensibility of these models companies is not to have the proprietary model that is better than another. That advantage asymptotes within a few years. They have a short window to create network effects on top of those models with workflow software, or security layers, or marketplaces, or 10 other methods. But that window closes. DeepSeek or the next DeepSeek, whoever it is, signals the closing of their window. These big dick swinging $500B projects, "I'm good for my $80B," or whatever, just got caught up in the competition with each other, ignoring the reality of the technology, misallocating VC money, shareholders money, and now trying to get taxpayer money. Once again, faster and smarter beats bigger. And look, if DeepSeek turns out to be not as great as we think now, don't worry, the next DeepSeek will be. And the next.

And here's the original three person unicorn essay.

@manav2906 @rrhoover ChatGPT Google Search Crunchbase Owler Product Hunt archives Reddit archives Articles from TechCrunch and others which mention competition Press release archives Twitter archives LinkedIn go take people to lunch and ask them

It’s possible the mobile tech window shortened the overall consumer tech window. At the time it felt like mobile was accelerating everything. It took incumbents too long to understand the internet tech window and they mostly died. The incumbents understood mobile after only three years.

RT @rrhoover: On building a consumer startup today by @jamescurrier. This is why I started Product Hunt 11+ years ago: To follow the trend…

For more on the consumer software winter we discuss it here. It’s thankfully over. @joshelman was the one who first suggested to me in 2015 that the initial consumer software technology window might have closed. He turned out to be right.

@worlduano @jeremysliew Of course, anything is possible. I’m just trying to help you give yourself the highest probability of success.

@zebird0 @NFX You could give us that chart! Would be cool to see. I would expect it to be dominated by the power law of a few big outcomes from the beginning of the technology window. Like in the B2B software chart, the later companies were mostly $1-$5B, the earliers ones $300B-$1T

@alex_gurevich Agreed. The incumbents have big advantages in the race to acquire customers who are buying new AI capabilities. It will be hard for new startups, just like it was in the mobile technology window where my estimate is incumbents took 93% of the value at least.

@AndyLapsa Congratulations to the whole Stoke Team! Character and culture win everytime.

@NFX bar for investing in startups is very high. We see 10,000 companies per year and invest in 25. More on what it takes in the above essay. If you have it, contact us.