> It was a strange feeling the first few times executing liquid allocation steps produced by an entity that wasn’t myself or a colleague. I felt like a liquid-handling robot running an experiment for Claude.
I can really relate to this feeling! Although my experience with letting Claude tell me what to do in the lab has been pretty hit and miss (I'm asking it more about equipment design and measurement), and I'm trying to build up a better intuition for how to balance trusting it vs. digging into it's suggestions.
I'm not sure what your setup is exactly but typically Claude code has given me dumb suggestions either because 1)it didn't understand the physical situation and limitations, or 2) didn't have context on the data I am collecting or troubleshooting. In 1) the programmatic definition of the physical experimental setup really really helps and Claude/codex can help you built that, and 2) this is the harder part and sometimes only starts working after consistent use.
I really like this - I’m thinking about how useful that Claude session would be for a new lab member to catch up on the work being done. To craft exact protocols that can be learned quickly. I have also found that a lot of where is a good place to start questions can be answered by a web search from Claude and then going from there. Thanks for the piece!
A lot of people still don’t seem to grasp that you need to use both together. You can’t just hand everything over to AI, some things still have to be explicitly programmed.
> It was a strange feeling the first few times executing liquid allocation steps produced by an entity that wasn’t myself or a colleague. I felt like a liquid-handling robot running an experiment for Claude.
I can really relate to this feeling! Although my experience with letting Claude tell me what to do in the lab has been pretty hit and miss (I'm asking it more about equipment design and measurement), and I'm trying to build up a better intuition for how to balance trusting it vs. digging into it's suggestions.
Cool to hear!
I'm not sure what your setup is exactly but typically Claude code has given me dumb suggestions either because 1)it didn't understand the physical situation and limitations, or 2) didn't have context on the data I am collecting or troubleshooting. In 1) the programmatic definition of the physical experimental setup really really helps and Claude/codex can help you built that, and 2) this is the harder part and sometimes only starts working after consistent use.
I really like this - I’m thinking about how useful that Claude session would be for a new lab member to catch up on the work being done. To craft exact protocols that can be learned quickly. I have also found that a lot of where is a good place to start questions can be answered by a web search from Claude and then going from there. Thanks for the piece!
A lot of people still don’t seem to grasp that you need to use both together. You can’t just hand everything over to AI, some things still have to be explicitly programmed.
Would you like to test our our agent, Elnora? https://elnora.ai/