fun frequent convo I've been having with founders in the a16z speedrun program we're running in LA for the next 12-weeks:
the clarity of a startup's story means clarity in their strategy. The story is the strategy.
On the other hand, a confusing, rambling story means that your approach is likely complex, incoherent, and unattractive. If you've ever had someone pitch you a product idea and then needed to ask question after question to understand what they're doing, you know what I mean.
This is a skill. It takes tremendous time and effort to pare back your startup's core ideas into a short series of truths, with strong local connections in between. The process of paring down the idea requires hardcore prioritization -- knowing what parts of the idea are core and must be done right away, and what parts are irrelevant, or can be done in the future. A clear narrative forces clarity in strategic tradeoffs.
It takes no skill to have massive complexity -- we can all brainstorm a ton of ideas, throw all the exciting trendy technologies into our idea, and make a go for it. But these messy stories reflect strategic indecision and avoidance. And the resulting unclear strategy will repel potential customers, investors, employees, and partners. They don't know who you are, or what you're really doing, because you don't either.
the best stories are a form of lossy compression, where they leave out 90% of the operational detail, but still hit the emotional and strategic truth. They ring true as soon as the words leave your mouth. And not because they are obvious statements -- after all, we get tired of hearing the same things -- but rather because they make observations on the reality of the world new way that cuts through the noise. They contain a "secret" that's surprising and compelling. The best narratives tap into feelings, not just logic.
the good news is that you can often figure this all out as you go. Founders and product builders are often are guided by the intuition of what they want to see in the world. Intuition might guide you towards a solid product, even if the initial story is a bit messy. And as you understand your customers' real world use cases, plus how your own customers describe your product, it can get easier to compress down all your ideas/observations/mistakes into a simpler expression of the idea.
London is the second-best city in the world to start an AI startup (and could be #1, in my opinion):
- Abundant early-stage capital
- High density of exceptional AI talent
- Significantly cheaper and more loyal talent compared to the Bay Area
- Pro-growth government and regulatory environment
Raising from a West Coast VC and then establishing your main office in London is an emerging, proven playbook I've seen repeatedly.
Perplexity CEO Aravind Srinivas says restrictions on chip imports to China are forcing them to innovate efficient solutions to AI model training, with DeepSeek trained on only 2048 H800 GPUs, making them the equivalent of DOGE for AI
I have historically been an Elon fan, and I think a lot of the anti Elon media has been unfair and/or misleading. He has created a lot of good for society.
But this story where he lied about being a "top 20 gamer" where he had someone leveling his account is hitting me hard.
As a gamer, you have a lot of pride in your rank/level. Trying to fake that while also being the richest guy in the world is just so absurdly unnecessary.
I would say it also makes me wonder, if he is willing to lie about this, what else is he willing to lie about.
It might seem trivial to many, but for the other people out there who have been competitive gamers, you know exactly what I'm talking about.
* Language is low bandwidth: less than 12 bytes/second. A person can read 270 words/minutes, or 4.5 words/second, which is 12 bytes/s (assuming 2 bytes per token and 0.75 words per token). A modern LLM is typically trained with 1x10^13 two-byte tokens, which is 2x10^13 bytes. This would take about 100,000 years for a person to read (at 12 hours a day).
* Vision is much higher bandwidth: about 20MB/s. Each of the two optical nerves has 1 million nerve fibers, each carrying about 10 bytes per second. A 4 year-old child has been awake a total 16,000 hours, which translates into 1x10^15 bytes.
In other words:
- The data bandwidth of visual perception is roughly 16 million times higher than the data bandwidth of written (or spoken) language.
- In a mere 4 years, a child has seen 50 times more data than the biggest LLMs trained on all the text publicly available on the internet.
This tells us three things:
1. Yes, text is redundant, and visual signals in the optical nerves are even more redundant (despite being 100x compressed versions of the photoreceptor outputs in the retina). But redundancy in data is *precisely* what we need for Self-Supervised Learning to capture the structure of the data. The more redundancy, the better for SSL.
2. Most of human knowledge (and almost all of animal knowledge) comes from our sensory experience of the physical world. Language is the icing on the cake. We need the cake to support the icing.
3. There is *absolutely no way in hell* we will ever reach human-level AI without getting machines to learn from high-bandwidth sensory inputs, such as vision.
Yes, humans can get smart without vision, even pretty smart without vision and audition. But not without touch. Touch is pretty high bandwidth, too.
@andrewchen Well said. The cross platform pay is essential. For use cases like learning for example, VR has the wow factor but pc (pixel streaming) + mobile will still give you a much. Wetter experience than video. Until VR picks up
muslim and arab (especially palestinian) colleagues in the tech community i've spoken with feel uncomfortable speaking about their recent experiences, often out of fear of retaliation and damaged career prospects.
our industry should be united in our support of these colleagues; it is an atrocious time. i continue to hope for a real and lasting peace, and that in the meantime we can treat each other with empathy.
I propose @JeffBezos as US President.
◆ Very balanced and nuanced in his speech and positions.
◆ Son of a Cuban immigrant. Understands the value of freedom and the American Dream.
◆ Knows how to scale massive companies as a collection of services with well-defined API boundaries, with a very large number of employees.
◆ He’s accomplished virtually everything possible in business, this is his ultimate test: public service. We forgive Amazon SimpleDB and Fire Phone.
◆ His muscles will intimidate and deter our enemies at home and abroad.
It was reported, based on scraped Amazon data, that Quest 2 is outselling Quest 3 by more than 2:1 this holiday season. That seems a healthy spread between an entry level and mid-range product. I am very glad they were able to keep Q2 in production — it is no longer an optimal design point today, but it is still plenty good, and it would have been disastrous for the VR market to move the entry point all the way up to the Q3 price point.
I remain unconvinced that mixed reality applications are any kind of an engine for increasing headset sales. High quality pass through is great, but I just don’t see applications built around integrating rendering with your real world environment as any kind of a killer app. I consider it interesting and challenging technology looking for a justification.
The power of VR is to replace your environment with something much better, not to hang a virtual screen in your real environment. In all the highly produced videos demonstrating the MR future, the environments are always stylish, clean, and spacious. That is not representative of the real world user base.
There is certainly some value in the efforts, but I have always thought there was much more low hanging fruit to be grabbed first.