πŸ‡―πŸ‡΄ Ben Little [they/them]

πŸ‡―πŸ‡΄ Ben Little [they/them]

@benlittle.dev

PA ➑️ BoCo.LoLa πŸ³οΈβ€πŸŒˆ Queer β™Ώ Disabled πŸ’» Chronically Unemployed Hacker πŸ‡―πŸ‡΄ Look to Palestine

103 followers82 following337 posts

Every computer system should bring people closer together. Shoving tech in between people who are already in effective communication makes about as much sense an hiring an interpreter to stand between two people who are speaking the same language. Unless of course, the goal is to disrupt them.

The thing that I find hardest to wrap my head around when I see speaking positively of coding agents: Thinking a tool that saves maybe 5 minutes of DIY coding every once in a while is worth TRILLIONS of dollars. And you get less comprehension and worse recall of your own decisions.

πŸ‡―πŸ‡΄ Ben Little [they/them] reposted

πŸ‡―πŸ‡΄ Ben Little [they/them] reposted

The human brain, with 86 billion neurons and a million billion connections between them, runs on 20 watts, less than a lightbulb. Think we may be going down the wrong road believing data centres, needing megawatts or gigawatts of power, with all the environmental & climate costs, are the way to go.

Linux is dead in the next few years. You cannot shove 2M lines of C into ring 0 and still have security. However GNU/Hurd has basically no adopters, and I don't know of any attempt to build on L4. The entire computing industry and everything that depends on it is in for an existential shock.

How can you know something is correct if you can't even articulate what it is supposed to do? How can you know when it is appropriate to use a certain tool if you don't understand its limitations? This is why a tool that "does everything" is actually a tool that does nothing.

I don't see much real discourse on the fundamental problem of machine learning: computational complexity. We can burn 100x as much oil, steal 1000x as much data, and still end up with a model that is only marginally better than what we have now. Changes in architecture cannot change the math.