The move to the cloud meant that infra costs started to change daily rather than quarterly. The same will happen to dev costs with the adoption of coding agents.
A staple of tech interviews is to dig into common abstractions, but the starting level keeps rising, from
- high-level language down to assembly
- GC down to memory management
- frameworks down to vanilla
Will the next phase ask candidates to leave the chat interface and actually type in the code?
A corollary to Arthur C. Clarke’s notion that advanced technology is indistinguishable from magic is that technological breakthroughs go through a “magic” phase until users develop mental models (accurate or not) of what’s going on under the hood. LLMs are currently in this phase.
@docmilanfar Whether AI is ML or vice versa, he’s right about the combination of Stats, CS, and Econ being the trifecta to master the next generation of problems.
The amount of coding LLMs can do is impressive, but so is the amount of drudgery developers treated as the norm. There was a lot of low-lying fruit in dev workflows that has been waiting to be automated for years.
Watched Polar Express, aka Uncanny Valley Snowpiercer, with the family yesterday. The animation hasn’t aged well, but at least the story is embellished and tonally unbalanced.
If future generations watch action movies where villains have only robot goons, will they look back on the human cannon fodder of, say, James Bond films as barbaric?
Whitehead said that “Civilization advances by extending the number of important operations which we can perform without thinking about them.”
He was talking about math, but the same could be said of software systems, most recently LLMs. Black boxes are a blessing and a curse.
The children’s books I read to my kids seem to fall into two broad categories: Those that are a pleasure to reread, and those that become more grating each time. Few lie between these two extremes.