how to do innovation

2025-04-01

you hear a lot of pieces of advice on how to achieve startup innovation, but I always struggled to put it in a big picture.

what you are trying to maximize is the generation of high-value hypothesis.

this means they are contrarian & important - when you test them and they turn out to be right, you win.

they are created through a cycle of exploring & building.

you gain new knowledge, come up with a take, and test it by building a prototype.

if it works, you get to leave the cycle for the next 7 years and build a startup.

otherwise, try again.

in order to be an effective "cycle explorer", there are three things you have to do:

  1. choose the right fields
  2. explore efficiently
  3. see things the right way
  4. be able to ship quickly and show the world

you should also be in the right environment (essay on that coming soon).

choose the right fields by fields I mean two separate things: a) a technology (ML, neurotech, AR) b) a domain (healthcare, cybernetics, education)

most people start with a technology, by either studying it or self-teaching out of interest.

"what do you think are going to be the most important fields? where will value emerge?"

you need to look into the future - it takes time to get cracked at the edge of a technology.

if you look at something, that will be important in 4 years, you will be exactly right.

I leave it to you to figure out how to learn efficiently.

besides specializing in one field, I would look out for overlapping areas you could learn using the 80/20 rule - like a T-shaped build.

it allows you to create things at the intersections others would have missed

a domain is a specific field / application which you will revolutionize or create using your technology.

I like to avoid direct competition by asking "what important fields are underserved?".

I find researching and building niche things simply more interesting and impactful than playing the business game.

when looking at innovation, I like to divide it into two categories:

  • pain killers - those that simply solve an existing problem (like Uber)
  • new value - those that create new desires (like Apple).

the second one is also called a behavior explosion.

you create new capabilities, experiences, or change how we interact with the world.

I think it is inherently harder and rarer.

there is no easy way to validate, you have to convince your users, that they even want this in the first place.

you need to build up taste and intuition for such things, you can't just methodically go around looking which problems you can solve, that weren't yet.

you need to predict how humans and the world will behave in the future.

good thing is, that if successful, you create a new ecosystem, not just a solution

look at how personal computers became a platform for even more startups and thus allowed this branch of human knowledge to grow even further.

a few more tips on the kind of ideas you should pursue:

  • "dogfood your own product" - only work on things your obsessively curious about - someone who is just after the money will not be able to compete - both on product and drive
  • the best ones are those that are tedious but you still love - cultivate weird tastes
  • TODO: add more

I personally started with ML when it wasn't popular yet (hit!).

also recently started with neurotech, because I believe it will be another hit in 4 years.

I broke down what domains could create a behavior explosion

a few items from this list as an example

  • memory augmentation
  • identity exploration
  • new forms of personal learning
  • cognitive enhancement
  • affective computing

from them I picked the 2-3 domains I am curious about, best positioned to work on, and believe will be very important.

once you have that, you can start exploring.

though remember that as new information comes in, you will want to adjust or even pivot completely.

see things the right way

you have learn to think in a certain way.

learn to observe: become obsessive about studying behavior: humans, systems, organizations, AIs...

learn to notice: train yourself to notice friction and inefficiencies instead of accepting the status-quo. doing so requires creative thinking and existing knowledge you can draw connections to or pull mental models from.

learn to form hypothesis: takes that are important, precise and testable -"why did previous attempts fail, and how are you different?"

explore efficiently get an overview over the field as whole: map, history, key figures, etc.

then simply follow your curiosity and sense for potential value into subtopics.

identify frontiers and open questions

  • current constraints & issues
  • existing approaches (or workarounds)
  • current developments & future hopes

have an extra look at power users and edge-cases.

because of your amazing pattern recognition skill, you notice that there is something wrong and you can fix it, because you have a unique insight most people missed.

a hypothesis is born

test quickly you need to be a builder.

you just have to.

after you tell me about a cool idea, I expect you to have a working prototype at the end of the week.

you need to have this sense of urgency and bias towards action.

I love craftsmanship in engineers.

they have their tools and processes nailed down.

for developers this means have a techstack you are insanely confident & quick in - have boilerplates ready

being a generalist and having worked on many different kinds of things obviously helps.

develop great problem-solving skills.

yes, this sounds vague, but there are books written about this for a reason.

next did you just hit gold?

amazing, now comes the step of execution.

there are books written about that, once the time comes I am sure, that you will quickly learn and succeed to:

  • raise money
  • share your vision
  • grow a team
  • build your product
  • distribute
  • defend and constantly innovate

Good luck!

☘