Two winters ago my flight out of Denver sat on the tarmac for three hours with the seatbelt sign on. My laptop battery was at sixty percent, the airline wifi wanted twenty dollars, and I had one video cached locally: Rich Hickey’s “Simple Made Easy” from Strange Loop 2011. I had watched it once in 2012 and remembered almost nothing. This time, trapped, no phone signal, no way to skim, I watched it properly. By the time we taxied back to the gate I had two pages of notes and a completely different idea of what the word “simple” means in a system.
Three hours of enforced attention on a decade-old talk did more for my engineering judgment than the previous three months of blog posts. That is the whole argument I want to make here.
Thirty minutes, three years compressed
A good conference talk is a compression artifact. Somebody spends three years fighting a problem, gets it down to thirty minutes, and stands on a stage telling you where the bodies are buried. Bret Victor’s “Inventing on Principle” (2012) is an hour of one person’s decade of thinking about how programmers interact with what they make. You can absorb the thesis over lunch. Buying that thesis any other way costs years.
Compare that to most written material. Articles have to hedge, cite, and anticipate objections. Talks are allowed to be an argument. The speaker picks a position, defends it out loud, and you get to watch a mind work in real time, pauses and stumbles included. That is how engineering knowledge actually transfers: one practitioner telling you what they learned the hard way.
The compression is also why talks hold up. Hickey’s “Simple Made Easy” is from 2011 and the ideas arrive unspoiled by fifteen years of subsequent commentary.
The debate is the good part
Here is the part I did not appreciate when I was younger. Old talks preserve the arguments of their era, and the arguments are the education.
Watch Pete Hunt introduce React at JSConf in 2013 and you can hear exactly which parts the audience pushed back on, and why the virtual DOM sounded reckless to people who had spent years learning to minimize DOM writes. Watch Douglas Crockford’s JavaScript talks from 2009 and you learn what the language looked like to a sharp critic before ES6 fixed half his complaints. Gary Bernhardt’s “Wat” (2012) is twenty minutes of comedy that teaches you more about JavaScript’s type coercion than most type-theory primers, and his “The Birth & Death of JavaScript” (2014) jokes its way to something like WebAssembly a couple of years before WebAssembly shipped.
When you only read today’s docs, you inherit conclusions. When you watch the talks, you see the deliberation that produced them, and deliberation is the transferable skill. Ten years from now someone on your team will propose something that sounds exactly as strange as the virtual DOM sounded in 2013, and you will have some calibration for how those bets usually turn out.
How I watch
I keep a rough system, and it is worth sharing because watching passively is nearly useless.
I watch at normal speed the first time and 1.5x only on rewatch. I keep a text file open and write down every claim that surprises me, then pick one and check it against reality during the week.
I follow speakers more than conferences. A speaker who was good in 2013 is generally still worth hearing in 2023. And I rewatch the great ones every couple of years. “Inventing on Principle” hits differently at thirty-five than at twenty-five, because I have more failed projects to measure it against.
Where to find all this: Confreaks is the archive I keep coming back to, with talks filed by event and speaker going back to the era I have been describing. Search a topic you think you already understand, sort by age, and watch something from 2011.
The education that costs nothing
A computer science degree costs four years and a pile of money. A curated hundred hours of conference talks costs an internet connection and the discipline to watch with a notebook. Both have gaps. But for the layer where most of us work, distributed systems behavior, language design trade-offs, performance, concurrency models, and the craft of reading code, the talks are the better teacher, because the people giving them are the people who built the thing.
My flight eventually left, three hours late, and I spent it reviewing my notes instead of a movie. The bug I had been avoiding in Denver got fixed the following Monday in about forty lines, using an idea about separating identity from state that I stole directly from a 2011 talk. The best CS education on earth is sitting in public archives, free, waiting for your next delayed flight.
