Crypto’s all about being open – creating new money legos (protocols anyone can plug together to create bigger things).
@zora + @base.base.eth are perfect examples – building blocks for creators, traders, and innovators to grow on. We’re putting the pieces together right now at clankfun.
Thanks to @jacob + @dee + @jesse.base.eth for the bricks 🧱
current claude code loop:
- give it the spec as a PM + any technical constraints. pass in documentation + mention specific files if relevant. ask it to come up with a plan
- read the plan, 60% of the time it's too complicated, ask it to simplify. read the next plan, usually approve it at this point.
- do a set of calisthenics or read something whilst waiting 2 minutes whilst keeping an eye on it. sometimes I have to interrupt execution because it starts doing something stupid.
- brief review of the code + test the feature
- iterate to fix bugs
- sense check if any feedback I gave was structural and likely to re-occur, then add the context to CLAUDE<.>md
- commit the changes as a checkpoint
one thing that serves me well when building is that I'm incredibly lazy with the early versions. I'll ignore pretty much everything outside of the highest-risk/uncertainty area in the product and build a throwaway prototype to derisk it.
for example: current client wants to build an HR operating system for fast-growing scale-ups that uses a voice agent. my first MVP is literally throwing a bunch of manually collected data into the system prompt and getting the client to test it. it took me half an hour but we now know the limits of the tech for their use case.
the next version is going to slightly simplify gathering the data and make it about 50% less manual so we can test it 5 times instead of once. I aim for at least one such test per day of building and make sure I test the most uncertain stuff first, and as a result progress tends to be pretty fast and smooth.
llm-powered dev is so different to traditional dev that if it was a job it would have a totally different set of competencies, maybe closer to "software architect" than "software engineer"
the unique skill needed to adjust is a sort of acceptance that you previous skills are becoming fast irrelevant and the ability to quickly accept that and move past that + start learning the new stuff
if you aren't going through that right now, you're probably coping, and it's better to just move past that and start learning the new skills and stay relevant for the long-term
if I were to coin a project again, here's how I would do it to avoid creating bad incentives:
1. build a product first, without a coin.
2. iterate to reach PMF without anyone suspecting there to be an airdrop. people are only using the product because the product is useful to them.
3. do a community fundraise with an ICO
I believe that anything else is going to make PMF harder to reach - just as it is harder to see the bottom of the sea in murky water, so is the right product shape amidst a user base of speculators.
main reason I'm spending less time on X is the sheer amount of fake content. the only reason that's not the case right now on farcaster is because there is not enough attention on the platform... yet.
the incentive structure in coined projects pits the average "community member" against the ultimate success of the project.
why? real investors will buy early and typically forget for months. but the loudest members in the first few weeks are incentivised to see the coin pump in the shortest time frame possible. thus, any new project will see the most vocal of the community members suggest: narrative, marketing, shilling, partnerships, paying DEX, getting on exchanges, and so on - anything other than building actual value into the product.
it requires a lot of resolve not to lean into this as a founder. and most will lean in, even just to compromise with a substantial chunk of their early community.
this is why we end up with a distinct brand of hollow "vaporware" in crypto - flashy websites, complex terminology, chapters of whitepapers, social media interns hiding barely functional, barely useful, barely interesting products.
the problem with traders larping as users is that it's a botched signal for the founder/builder.
you read the silicon valley product canon and learned that the most important thing is to listen to users. but traders will tell you that to make them happy, you have to spend all of your money and energy to create a short-term pump of your token (so that they can exit your token and your telegram).
if you mistake traders of your coin for users of your product, you might actually listen to them and make such terrible decisions.
weird thing about building in web3 is that you get a lot of traders pretending to be users. users care about what you're building. traders work hard to pretend that they care, because doing that looks bullish in your telegram.
when they have an open position, they will act as if what you're building is the most important thing they've ever heard about. obsessed fanatics. as soon as your token's price stagnates, or they took profit, they instantly disappear.
