undefined | Better HN

0 pointseastbound7mo ago0 comments

Today I’ve compiled a few thousand classes of Javadocs in .978 second. I was so impressed, with a build over 2 minutes, each byte of code we write takes a second to execute, computing is actually lightening fast, just now when it’s awfully written.

Time of executing bytecode << REST APIs << launching a full JVM for each file you want to compile << launching an LLM to call an API (each << is above x10).

0 comments

kylemaxwell7mo ago

The point is that you call the LLM to generate the code that lets you talk to the API, rather than writing that glue code yourself. Not that you call the LLM to talk to that API every time.

simonw7mo ago

Exactly.

j / k navigate · click thread line to collapse

0 pointseastbound7mo ago0 comments

Time of executing bytecode << REST APIs << launching a full JVM for each file you want to compile << launching an LLM to call an API (each << is above x10).

0 comments

kylemaxwell7mo ago

The point is that you call the LLM to generate the code that lets you talk to the API, rather than writing that glue code yourself. Not that you call the LLM to talk to that API every time.

simonw7mo ago

Exactly.

j / k navigate · click thread line to collapse