LLMs allow JIT situated answers, for queries and also glue code.
- LLMs allow JIT situated answers, for queries and also glue code.
- Before, the answer to your query had to already exist before you searched.
- That content was produced by a stochastic distributed swarming process of content creators.
- People out in the world had to think that maybe someone would find such an answer valuable, and then invest time and effort to write it and get it indexed.
- It worked surprisingly well, but with a lag, and also with answers that were one-size-fits-all and not situated to your particular needs.
- But now LLMs can give perfectly situated answers on demand, JIT.
- This process can fill fractally more tight niches than the previous one.
- Similarly, having interoperability between two programs had to have pre-existing glue code.
- Someone else would have had to think to write interoperability glue code for that precise pair of things you cared about.
- Glue code requires being meticulous and is all downside. I
- t is not fun.
- They'd have to care quite a bit, to make it robust and maintain it as they change.
- That process was stochastic; for many pairs of things that could be glued together, you'd just have to do it yourself.
- Now LLMs can distill glue code JIT.