David Rein
@idavidrein.bsky.social
sentio ergo sum. developing the science of evals at METR. prev NYU, cohere
Hey, study author here—I think this is an overgeneralization.
We find that *experienced, open-source developers working on projects they’re highly familiar with* are slowed down. This is consistent with many developers being sped up often, e.g. when writing one-off scripts
We find that *experienced, open-source developers working on projects they’re highly familiar with* are slowed down. This is consistent with many developers being sped up often, e.g. when writing one-off scripts
July 11, 2025 at 12:30 AM
Hey, study author here—I think this is an overgeneralization.
We find that *experienced, open-source developers working on projects they’re highly familiar with* are slowed down. This is consistent with many developers being sped up often, e.g. when writing one-off scripts
We find that *experienced, open-source developers working on projects they’re highly familiar with* are slowed down. This is consistent with many developers being sped up often, e.g. when writing one-off scripts
had a similar experience in africa—zebra and wildebeest were just chilling like right next to lions
November 28, 2024 at 9:44 PM
had a similar experience in africa—zebra and wildebeest were just chilling like right next to lions
If interpretability gave us as much info as transparent cases did, we’d be a lot further along than we currently are
April 27, 2023 at 4:35 AM
If interpretability gave us as much info as transparent cases did, we’d be a lot further along than we currently are
omg so true, thanks for letting me know!!
April 27, 2023 at 4:33 AM
omg so true, thanks for letting me know!!