Cars turned us—one of the best species in long distance running into couch potatoes.
Now llms are attacking our brains and making us stupid and insane. A species of slopheads if you will.
Cars turned us—one of the best species in long distance running into couch potatoes.
Now llms are attacking our brains and making us stupid and insane. A species of slopheads if you will.
Can’t go wrong with offline and open-source!
Locking in in this context means that you’re relying on an LLM for your problem solving - atrophying your skills in the meantime, making you dependent on a model. So you can indeed go wrong with open weight abd self hosted ones. At least thats my understanding of this.
Here’s the scary part… The skill of developers has been for about 20 years: “look it up on stack overflow. Find a similar problem. Fix it to suit my environment.”
But… No one is posting to stack overflow anymore. So LLMs have effectively become stack overflow.
I try to not let the skill atrophy by doing it the old way. Can’t. The well is running dry.
What I will say is that LLMs make parsing logs a lot easier. So, doing things the old old fashioned way is still in the cards.
It’s weird out here.
Remember when Google gave relevant search results? Pepperidge Farm remembers….
it wasn’t what i was getting at, but that’s also true. the model requires a certain setup to be effective, so now you’re locked into that. the model does things a certain way, so now you’re locked into that. nobody reads the code it produces, so now you’re locked into that.
all the while every other way of doing things disappears from your mind.
considering the massive accidents of models deleting codebases and production db’s, the fact that the tool may be open source doesn’t really help.
besides, i don’t know any open source models. i know of open-weight models, but i’ve yet to see anyone share the training regime and source data for an even vaguely effective model.