Post
Ethan Mollick
@emollick.bsky.social
Using Gemini 4 makes you realize how many other models must be distilling Claude/GPT (or using a similar training pipeline). It feels quite different than other models.
2:16 PM · Oct 13, 2026
4 reposts
124 likes
4 saves
cafkafk @cafkafk.bsky.social
· 10h
this but doing your own models, like a lot of things I thought were "llm" isms were really just claude/gpt isms
Jo Michell
 @jomichell.bsky.social
· 10h
In what way?
Gary Longsine 📱 @arbiteriapetus.bsky.social
· 7h
As one tiny example, Gemini doesn't share the "verbal tics" like overuse of the phrase "load bearing" that Claude displays. www.reddit.com/r/claudexplo...
From the claudexplorers community on Reddit: worked with claude to make a song about how it wont stop saying load bearing
Explore this post and more from the claudexplorers community
www.reddit.com
Evan Blake @evanprompts.bsky.social
· 9h
Same with Gemini 4 on rewrites, it doesn't reach for the phrasing the others all share. Makes me think half my prompts were tuned to Claude/GPT habits
James Greene @virdesi.bsky.social
· 10h
Considering ChatGPT is being ranked lower than Muse, will OpenAI distill Gemini 4 to figure out Google’s training pipeline then copy it without acknowledging what it did?