Talking to clouds

Most developers are not assessed by their ability to be innovative or creative. The standard ‘AI will just make everyone faster’ trope misses the mark by assuming every developer is also creatively original (far from true). How the labor market will adjust to this is unclear.

How many thoughts have you had in the last minute? How many are in parallel? How many are visual? How many are happy/sad? How many do you think the person next to you had? Have you ever wondered or asked?

With AI we are self-selecting out intelligence as an important genetic trait by making it cheap. Being hot, healthy, and happy will gain in its place.

Authenticity is a game of scarcity.
Quality is a game of abundance.

It makes sense to play games where quality is the driving force without relying on authenticity.

You will know the AI hype phase is over when people stop pitching AI as magic and identifying it using the sparkle icon. We have the power to simply set proper expectations like a real product.

Being able to access and parse internal monologue strikes me as the fundamental unlock of humanity.

Exploiting information asymmetry is the foundation that many industries are built on. It doesn’t get enough attention that even though AI raises up the capabilities of the middle majority, it also dramatically flattens the distribution. Most knowledge workers will end up having nearly indistinguishable skills except those outliers 2 or 3 standard deviations from normal. Those that do not use AI will be strongly disadvantaged, while those that deeply implement custom processes will gain superpowers. We all have access to deep knowledge now, imagination and execution (speed x quality) are the leverage points.

Reviews are likely more noise than signal. Consider the people whose taste and context you respect most. Now how many of those people do you think take the time to write public product reviews? Likely not many.

We can fix some of this with more empirical takes made by AI.

I think AI will create a new type of ‘late mover’ companies. Let big companies do the expensive work of innovation and business model optimization. Then just copy that and add a thin layer of niche-specific customization on top. Copying established products in high fidelity will become very cheap, letting indie founders copy a product group in homeostasis, but with a tangential twist. The key is enabling companies to do a frictionless swap from whatever they currently use.

We are deeply hindered in our ability to advance as a species by not having some form of human-to-human API. AIs will have this and that is an exponential unlock.

Rudimentary semantics will be our downfall.

History is written by the victor
History is written by search engines
History is written by model providers
History is written by the human feedback in RLHF
History is written by ???

(??? seems like the trend to invest in)

AI is a new tool, one that offers leverage. New sources of leverage increase available optionality. Additional optionality increases the scale of both good and bad outcomes. More is not better or worse by default. Thankfully, history has shown tools to be awesome at every turn.

It is really annoying that webpages don’t have proper metadata. How often is the page expected to change? What is the basic content type. When was the last time it was updated? What is a paragraph (or 2) summary of the contents (what, who, how much)?

I have never seen so many people building single-point dots, with no idea of whether line will connect them in the future. We need to make imagination cool again, not just ambition.