Dan Ratner
@danalert.bsky.social
technologist | writer | humanist
danratner.com and fka @danalert
danratner.com and fka @danalert
I’ve been coding with Claude and Deepseek. Deepseek gets the job done-good, clean, compiling golang code. But Claude has the edge for robustness like error checking.
February 7, 2025 at 3:51 PM
I’ve been coding with Claude and Deepseek. Deepseek gets the job done-good, clean, compiling golang code. But Claude has the edge for robustness like error checking.
It drives me crazy when people try to assess LLMs based on what they know. First, they were interfaces between "human" and "computer". Now they can do some reasoning,. But they can also be given or retrieve data in different ways (like RAG servers.) They were never meant to be search engines.
January 31, 2025 at 3:34 AM
It drives me crazy when people try to assess LLMs based on what they know. First, they were interfaces between "human" and "computer". Now they can do some reasoning,. But they can also be given or retrieve data in different ways (like RAG servers.) They were never meant to be search engines.
After playing with a lot of the big LLMs for awhile I've found they have personalities. Deepseek is "insecure", Claude is "servile", and OpenAI is "bro".
January 31, 2025 at 3:19 AM
After playing with a lot of the big LLMs for awhile I've found they have personalities. Deepseek is "insecure", Claude is "servile", and OpenAI is "bro".
Reposted by Dan Ratner
We're looking for some controversial accounts to follow. Does anyone have suggestions?
January 12, 2025 at 10:25 PM
We're looking for some controversial accounts to follow. Does anyone have suggestions?