After years of studying crypto, developing websites or tools (for other people by the way) I decided to become a full time developer in this field and I'm all on my own in most projects. In other fields, there are certain ways of getting into the field (I was teaching AI and computer vision way before ChatGPT, and the roadmap is a little different now since we have to teach people how to use agents/chatbots as well and the whole thing hasn't been changed much) and I'm sure there is same thing for crypto as well.
My problem (and a lot of people like me honestly) is being too misguided and there is always an area of questions in our minds. However, I usually tell people who want to learn machine learning that we start at a topic such as linear algebra, then we get into useful algorithms, etc. Or AI in general I ask what people want to build, a simple classifier or a language model? And guide them through language processing, image processing, computer vision, etc.
So I am familiar with concepts such as DeFi, NFT, DEX, Stable Coin, etc. I even built multiple projects, but I still don't know where I can join all this knowledge together and make it into a good idea, project, product or even startup.
So, I am all ears about your guidance.
I'm all on my own in most projects... I still don't know where I can join all this knowledge together and make it into a good
idea, project, product or even startup.
I'm a solo developer too. I maintain a small proof-of-work coin built from Bitcoin Core's code, so here's what actually taught me
things, rather than more "keep learning":
1. Learn from Bitcoin Core's source code, not from summaries. Clone it, build it, and run it on regtest. Then follow one
transaction all the way through: the wallet creates it, the mempool checks it (
policy), a miner puts it in a block, and
validation accepts the block (
consensus). Once you understand the difference between policy and consensus, DeFi, DEXes and
everything else make much more sense, because they are all built on top of those two layers.
2. Change the code and run the tests. Core ships hundreds of unit tests and Python functional tests. Change something, see
what breaks, and work out why. That teaches you more than any roadmap could.
3. Security is most of the job. Whatever you build, expect people to try to break it. In one week, outside researchers
found seven issues in my project. One of them meant a single cheap transaction could stop a node producing blocks. Patching
quickly, writing a regression test and saying publicly what happened is part of the work. If you launch something, set up a bug
bounty early, even a small one.
4. Pick a problem you're actually interested in, then build something small that works and keep it running. Running a live
network, even a small one, teaches you things a side project never will: nodes falling out of sync, peers you can't reach,
releasing for several platforms, and users who lose coins when an address check is missing.
5. On working alone: KingsDen has a point, but you can get a long way on your own if you automate hard. Use scripted
builds, scripted deployments, tests that block a bad release, and yes, AI agents for the routine work. You still have to
understand every change you ship, though. AI will happily write code that compiles and is wrong, which is exactly why point 1
matters.
If you want to see what that looks like as a working project, I'm happy to share how mine is set up. PM me.
