i’ve been on youtube a long time. long enough that my early videos look like a different person made them, long enough to have shipped through every algorithm change and platform shift, long enough to have figured out, mostly by repetition, what actually works.
a lot of that transferred when i started building software products seriously. not the surface stuff. the deeper patterns. the ones about how long things should be, how often to ship, what teaching does to your skills, and what to do when something isn’t working.
these are the things i actually use.
three-hour courses used to kill. now people want one short video.
this is the shift that surprised me most.
a few years back i could put up a three-hour deep-dive on react or node and watch it do real numbers. people would sit through the whole thing. they’d comment on something that happened ninety minutes in. the long-form course format was the format.
today, a three-hour video is a tough sell. and to be honest, chopping the same course up into a series of ten-minute videos doesn’t really save it either. that was my first instinct (same content, smaller pieces), and it didn’t move the needle the way i expected.
what people actually want now is one short, focused video that answers a single question. not a course. not a series. not a playlist they’re supposed to commit to. one video, ten minutes or less, that solves the specific thing they came looking for.
i don’t think this is a sign of the audience getting worse. i think it’s a sign that the world got noisier and people learned to defend their attention. they don’t want to be enrolled in your course. they want the answer they came for, and then they want their time back.
what it taught me about software: ship one thing that does one thing. webutils does small web utilities. eyebreak reminds you to take a break. typesmith does one specific design task. each of those is the software equivalent of a single focused video. the person who needs that exact thing finds it, uses it, and is done. they don’t have to learn a platform. they don’t have to commit to an ecosystem. one question, one answer.
the impulse to bundle everything into a “v2” launch with five features is the same impulse that made me think a three-hour course was a flex. it isn’t. tight wins.
cadence beats trying to make a masterpiece
the thing that took me the longest to internalize on youtube: the people who actually win are the ones who keep showing up.
not the most talented. not the ones with the best gear. the ones who post the next thing, and the next thing, and the next thing, for years, while the talented people who agonize burn out and quit.
this is uncomfortable because it implies your taste isn’t the bottleneck. your output is.
it transferred straight to building products. the projects i have that are alive today are alive because i kept shipping into them. not because the first version was great. usually because i shipped a hundred small versions and the right design only became visible after the wrong designs got tried.
caveat: cadence beats perfection at a quality floor. shipping garbage on a schedule is not the lesson. the lesson is that good enough, regularly, compounds in a way that great, occasionally, doesn’t.
teaching made me a better developer than building ever did
the most underrated benefit of doing youtube for as long as i have: it has made me a much better developer than i would have been otherwise. and i don’t think i would have figured this out if i hadn’t been doing it for so long.
you can’t fake your way through a tutorial. the moment you sit down to record an explanation of how something works, you immediately discover every gap in your understanding. you thought you understood react context until you tried to explain it on camera and watched yourself flounder. you thought you knew how oauth flows worked until you had to draw the diagram for someone else.
every video is a forced honesty pass. either you actually understand the thing well enough to explain it cleanly, or you sit back down and learn it properly so you can.
the second-order effect is bigger. once you’ve taught a topic, you understand it from multiple angles. you’ve answered the dumb questions and the sharp questions and the questions you didn’t know to ask. the next time you reach for that pattern in your own code, you reach for it correctly, because you’ve been forced to think about it from every direction.
if you’re plateauing as a developer, start teaching. write tutorials. record explainers. answer questions on a forum. it doesn’t matter which surface. explaining it forces you to actually know it, and there’s no shortcut around that.
the audience is slow, and they will surprise you
you don’t learn what your audience wants by asking. you learn by shipping and watching what they do.
what they say they want and what they actually engage with are often different things. videos i thought would tank sometimes did the best. videos i was sure of sometimes landed nowhere. after enough cycles you stop predicting and start posting.
same shape with software. the feature i thought users would love sometimes gets ignored. the small thing i added on a whim becomes the thing they tell me about. i don’t try to predict anymore. i ship, watch, adjust.
and it takes longer than you think for an audience to form. you’re looking at a curve that doesn’t start to bend for months, sometimes years, and the only way to be on the right side of it is to still be making things when it starts to.
kill what isn’t working, even if you’ve put months into it
every long-running youtube channel has a graveyard of series and formats that didn’t work. the trap is sunk cost: “i’ve already made five episodes, i should keep going.” you shouldn’t. you should kill it and put that energy somewhere with a future.
the test that’s never failed me: when you imagine killing the project, do you feel lighter or emptier?
lighter means kill it. you’ve been carrying it. emptier means it still belongs to you and you should keep going.
heuristics that help:
- if you’re waiting for someone else to validate the idea before you ship the next phase, the project is probably already dead and you’re stalling
- if you keep adding scope to make the project more interesting to yourself, the original scope wasn’t pulling you anymore
- if every status update sounds like the previous status update with the dates changed, momentum is gone
kill it. free the slot. build the next thing.
energy is the budget, not time
the amateur model is “i have eight hours, what can i do with them.” the professional model is “i have so much creative energy per day, where am i spending it.”
youtube taught me this the hard way. recording, editing, writing, planning, replying to comments. these are all “work” but they pull from very different reservoirs. you can edit for six hours straight. you cannot write a script for six hours straight. trying anyway is how you blow out a saturday and have nothing usable on monday.
i now think about my workday in energy lanes. mornings are creative work: writing, designing, the architectural decisions. afternoons are mechanical work: code review, edits, configuration, the deterministic stuff. evenings are nothing. i protect them. the lab runs better when its operator does.
the agent fleet exists in part because of this. the autonomous profiles on the homelab handle the deterministic, low-energy work: ops cron, security scans, calendar sync, drafts. claude code handles the high-energy pairing. the hardware is in service of the same budget. energy, not time.
build with the lights on
the last one, and maybe the most important.
youtube taught me to be okay being seen while building. you ship videos before they’re perfect. you publish ideas you might be wrong about. you let people watch you change your mind. it’s exposing, and it’s also the thing that makes the work travel, because nobody talks about a polished outcome the way they talk about a process they got to watch unfold.
this site is the same idea applied to product work. the devlog is me working in public. the project pages document what’s under construction, not just what’s finished. the next article exists for the same reason.
the gallery shows you the painting. the lab shows you the painter. people remember the painter.
a long youtube run taught me to trust that. the lab is the same bet, made on a different surface.