Writing is fundamentally the transfer of information from your brain to my brain. If you have 1000 bits of semantic information you want to transfer, you can't give 300 bits of semantic information to an LLM and have it fill in the remaining 700, because it doesn't know what those 700 bits are. If it's able to guess those 700 bits correctly, then they aren't true semantic information, and you really only have 300 bits you want to transfer. You might as well transfer those bits to me directly, rather than having the LLM add on an extra superfluous 700 bits that I then have to filter out.
The value of writing isn’t always to communicate new information. It’s often to align everyone’s assumptions. For instance, when I say casually to a colleague or an agent “this change will require a db migration” they understand it’s to my teams primary application database. If I submit a design doc to a company wide review which database is changing is critical information. If the 300 bits were truly enough your agent or junior engineer would implement the wrong thing correctly as they often do.
Clarifying which database you're referring to is exactly the sort of thing I'm talking about when I say "new semantic information". It should be communicated, and you can't trust an LLM to choose the correct database, you need to specify that yourself.
What I'm talking about is if you add "database foo" to your prompt, the LLM may then add text describing what that database is, where it is, etc. But that's not new information, it (hopefully) already exists in your team's public docs, slack convos, etc. You should just say "database foo" directly to your reader, and if they want to learn more about that database, they can do that themselves, or you can give pointers to them based on what you consider important.
> say "new semantic information". It should be communicated, and you can't trust an LLM to choose the correct database, you need to specify that yourself.
But I can most of the time, and correct it if it chooses the wrong thing. That's the whole reason LLMs are faster. Its the reason we can give a paragraph prompt and get a kLOC PR back but only need to correct about 5% of it.
hatthew · · focus · HN ↗
Writing is fundamentally the transfer of information from your brain to my brain. If you have 1000 bits of semantic information you want to transfer, you can't give 300 bits of semantic information to an LLM and have it fill in the remaining 700, because it doesn't know what those 700 bits are. If it's able to guess those 700 bits correctly, then they aren't true semantic information, and you really only have 300 bits you want to transfer. You might as well transfer those bits to me directly, rather than having the LLM add on an extra superfluous 700 bits that I then have to filter out.
bbatha · · focus · HN ↗
hatthew · · focus · HN ↗
What I'm talking about is if you add "database foo" to your prompt, the LLM may then add text describing what that database is, where it is, etc. But that's not new information, it (hopefully) already exists in your team's public docs, slack convos, etc. You should just say "database foo" directly to your reader, and if they want to learn more about that database, they can do that themselves, or you can give pointers to them based on what you consider important.
bbatha · · focus · HN ↗
But I can most of the time, and correct it if it chooses the wrong thing. That's the whole reason LLMs are faster. Its the reason we can give a paragraph prompt and get a kLOC PR back but only need to correct about 5% of it.