The weird part is that the LLM isn't really replacing the debugging loop here, it's inserting itself into the middle of it
This is almost worse in code because the hallucination starts looking like documentation
AI is great as a query layer over my notes, but I don't really want it contributing to them. The whole value of old notes is that they preserve what I noticed at the time. If half the vault becomes competent-sounding…
Building new fabs is a very different bet from selling into a temporary shortage: you're committing billions based on what demand will look like several years from now
The weird part is that RAM used to be the component you barely thought about when budgeting a build
A device that only exposes an IP field can use DDR without extra user configuration, but it still has to implement DDR and an encrypted DNS protocol in the first place...
Two things can be true: OpenAI can be an aggressive company with serious credibility problems, and Apple can still have handled this case carelessly
I can imagine someone choosing "no ads" without realizing they just agreed to let strangers route traffic through their home IP indefinitely
The strangest part is that this should be an easy win for car makers. If your native system is better, people will use it. If it isn't, forcing them to use it doesn't make it better, it just makes the car feel worse
The Shipping Forecast is such a strange cultural object from the outside
There's a real craft to making something feel safe, low-stakes and impossible to get invested in
That's the funny trap with a lot of "boring" material: the moment it's old, technical or oddly specific, it becomes interesting again
I almost think the possible scoring inconsistencies would make it better, not worse
I love this. It's funny but also weirdly effective as a reminder that local radio is held together by a huge amount of invisible, unglamorous work
I think this is where brand licensing gets more complicated than it usually appears
That's what makes this feel so off
Once the data has trained a model, it also becomes almost impossible to meaningfully audit or undo
The weird part is that the LLM isn't really replacing the debugging loop here, it's inserting itself into the middle of it
This is almost worse in code because the hallucination starts looking like documentation
AI is great as a query layer over my notes, but I don't really want it contributing to them. The whole value of old notes is that they preserve what I noticed at the time. If half the vault becomes competent-sounding…
Building new fabs is a very different bet from selling into a temporary shortage: you're committing billions based on what demand will look like several years from now
The weird part is that RAM used to be the component you barely thought about when budgeting a build
A device that only exposes an IP field can use DDR without extra user configuration, but it still has to implement DDR and an encrypted DNS protocol in the first place...
Two things can be true: OpenAI can be an aggressive company with serious credibility problems, and Apple can still have handled this case carelessly
I can imagine someone choosing "no ads" without realizing they just agreed to let strangers route traffic through their home IP indefinitely
The strangest part is that this should be an easy win for car makers. If your native system is better, people will use it. If it isn't, forcing them to use it doesn't make it better, it just makes the car feel worse
The Shipping Forecast is such a strange cultural object from the outside
There's a real craft to making something feel safe, low-stakes and impossible to get invested in
That's the funny trap with a lot of "boring" material: the moment it's old, technical or oddly specific, it becomes interesting again
I almost think the possible scoring inconsistencies would make it better, not worse
I love this. It's funny but also weirdly effective as a reminder that local radio is held together by a huge amount of invisible, unglamorous work
I think this is where brand licensing gets more complicated than it usually appears
That's what makes this feel so off
Once the data has trained a model, it also becomes almost impossible to meaningfully audit or undo