It’s not rare for people to succeed despite being: - arrogant - volatile - low eq - not particularly intelligent - disorganized - unhealthy lifestyle - health issues (ones you can’t do anything about) But it’s extremely rare to succeed with the following attributes: - lazy - no ambition - focused on self-entertainment - only thinking about today or maybe tomorrow - strongly-held self-limiting beliefs If you are trying to “fix yourself” the items in the second list take higher priority. (Credit to Paul Graham for introducing this contrast)

I’m looking for a dev with the following: - Must be able to do UI / UX design if given a brand guideline and preexisting copy - Must be comfortable with php - Must be based in Asia - Must have used with Selenium / Playwright for website testing (or for sniping concert tickets/sneakers, I don’t care, but you MUST have used the stack). - Six day work week - Already using AI in your workflow. If a page design is extremely well specified, you should be able to have a same-day turnaround. Exact “years of experience” doesn’t matter, but the above requirements are a MUST and not something to learn on the job. DM me if this is you. I can’t make exceptions to any of the above points — I’m really sorry.

There should be at least one subject that you understand very deeply. People who don’t get to this point don’t learn what “deep understanding” looks like. They think that “understanding” a subject means having consumed the most popular content on the topic. Once you understand one subject deeply, you can re-apply that experience to learn another subject deeply. If you know one subject deeply, it’s easy to learn other topics to a shallow-medium level — but if you know multiple subjects at a shallow level, you don’t know how to learn one subject deeply. Furthermore, the major edge humans have over AI is the ability to develop a deep understanding of a subject — AI already has a shallow understanding of everything but struggles when it gets to the fine details.

This never fails to put a smile on my face.

I’m having the legendary @Jeyffre on THE NETWORK PODCAST this Friday. What should I ask him? Respond with a comment below this post!

Let me remind you for the 56th time that being a “monkey see, monkey do” coder will not get you places. You should actually learn the foundations of computer science. Most true innovation in computer science stopped decades ago. The “innovation” we see nowadays is simply re-arranging things that already existed + better hardware and faster networks. (Not an insult to innovators today, genuine innovation today takes skill this decade just as it did decades past). The way you stay “caught up” with the latest changes is by learning the foundations. - Rust is basically a mish-mash of C++ and functional programming - AI is just computer graphics, linear algebra, and a touch of calculus - Blockchain is just distributed compute with byzantine fault tolerance requirement - SVM, EVM, CairoVM, etc are just variations of the Von-Neumann architecture with minor changes. - ZK programming (Circom, Halo2, etc) is merely a cousin of logic programming. - ZK itself uses math and cryptography algorithms that have existed for a long time. If you are struggling to keep up, it’s not because you aren’t learning fast enough, it’s because you don’t understand the bricks that everything we call “computer science” nowadays is made up of. Metaphorically, if all you can do is stick frozen pizzas into the microwave, you will eventually be replaced. If you know how flour, yeast, cheese, etc interact with each other, then you can be a real chef and adapt even if ingredient availability changes.

Something I’ve been actively thinking about is “How do you teach someone how to write?” I’m bulling on using AI to learn how to code, but if AI is your writing instructor, it can only help you if your writing is really bad. Some thoughts in no particular order: 1) Direct mentorship is very effective if it can be sustained for multiple years. Some of the writers who worked for me for that long just astound me with how well they can write, and our only interactions are me giving feedback about how to improve an article. It worked a lot better than I thought. Eventually, they develop a second nature for recognizing substandard writing (even if subtle) and it’s amazing. This is not a scalable model however. 2) Writing well requires volume in the early days. Writing is like running. If you are a beginner, there is no such thing as “working hard” vs “working smart.” If you are overweight and you start running, you will see results regardless. “Working smart” only matters when you are approaching your genetic limits (which most people don’t). If you are a business owner, you can automatically improve the writings skills of your employees by replacing your standups with written standups, then giving feedback on the writing when it isn’t clear what the update is. The nice thing is, people are an expert on what they worked on, so they only have to practice expressing it. Volume, for an early writer, is the only thing that really matters. 3) Most people have an undertrained vocabulary. Consider this analogy: Even if you never “use” math you should still learn it. Even if you never “use” recursion or DSA, you should still learn them. They train you on how to break a large problem down to smaller ones, which is basically your job as an engineer. For writers, your job is also to break a larger idea into smaller ones. Since words are the indivisible units of ideas, you need to have a good sense of them. If you can articulate the difference between - “insinuation” and “implicature” - “generalization” and “abstraction” - “foresight” and “forecast” - “interpolation” and “induction” then you are automatically training yourself to express ideas precisely — which is the whole point of writing.

It’s amazing to me how busted education is, even in college. 1) You sit in a lecture where the teacher yaps the same thing for the dozenth time. He or she probably doesn’t have the advantage of special effects or animations to convey concepts more clearly (i.e. visual modality is nerfed). The teacher might not even be good in the first place. 2) The class moves at a one-size-fits all pace 3) Hours later, maybe days later, you do the homework. This is after you forgot everything you learned in class — assuming you learned something in class. 4) You probably cheat on your homework and don’t actually learn anything. If I designed a school, every student would sit in front of a desktop (with access to the internet cut off or extremely limited), and it would run a software that gives bite-sized lessons and a couple exercises to practice it. A proctor would wander around to make sure students are focused. Nobody takes homework home. When you are done, it’s over. It just flat-out does not make sense to me why we use teaching strategies from 400 years ago as if tech hasn’t changed since then. I understand this strategy doesn’t work for every subject (especially writing), but a lot of subjects, even liberal-art ones, can be taught in this manner. Also, the computers would run Arch Linux. Because I said so.

imo Jeff's biggest contribution was that he nailed ZK education with the perfect opening sequence: Sets, Groups, Fields, Homomorphisms, Pairings. this bit cannot be understated; absolute game changer. I say this as someone with a pure math grad with 2 quantitative masters. this unlock was a massive acceleration factor. i'd probably take a year instead of 2 weeks, without that correct sequencing. No one else simplified it like this; all ZK resources now reference this unlock.

In an earlier tweet, I recommended working on the hardest problem you can to gain respect, and then use that respect to get out of the “junior engineer” stage. But this doesn’t apply only to juniors. In my own work life: 1 - I became the youngest engineering manager at Yahoo by trying to solve real-time AI / AR back in 2017. We have SDKs for that nowadays, but back then it required significant optimizations due to limited hardware. I didn’t say “promote me, and I’ll solve it.” They basically begged me to manage more people after I showed I could deal with a problem they thought wasn’t solvable. 2 - RareSkills started to be taken very seriously after I cracked ZK education. Back in 2023, I could count on one hand the number of resources that made a serious (though incomplete) attempt to teach someone how to actually code a prover and verifier. Nowadays, there are a lot of ZK materials, but most of them are heavily inspired by the ZK book. It may seem obvious to study Sets, then Groups, then Fields, then Homomorphisms, then pairings, nowadays, but not even the moon math manual had such a clear knowledge graph (more than one Math PhD told me it was hard to understand, but I don’t mean this as an insult to the authors, putting all that material together is no small feat). It was ferociously hard to find a new way to explain ZK besides the useless “polynomial commitment -> linear PCP -> non-interactive proof” explanation that people didn’t actually understand anyway. —— When you are more advanced in your career, the temptation may be to just focus on what is convenient or profitable in the short term. I spent years dogged with the feeling that I should just get a high-paying technical job instead of writing blogs people read for free. But consider this: doing hard things can create more profit opportunities in the long run. For example, technical writers are normally paid very little, but we managed to turn it into a profitable business (and pay our writers 2-3x what they would earn elsewhere) simply because we proved we could do something hard in the field of technical writing. Do not underestimate the power of reputation and respect and do not try to shortcut your way there. The reason accomplishments gain you respect is not because of the accomplishment itself, but because of the implied effort you put in on the way there.