random dude

[random dude @nedjmi_]

@osanseviero Does it work with pdfs out of the box or you gotta convert them to images

[random dude @nedjmi_]

@GoogleAI Does it work with pdfs out of the box or you gotta convert them to images

[random dude @nedjmi_]

@burkeholland @svpino Why no o3 ??? It's cheaper than gemini 2.5 pro!!!

[random dude @nedjmi_]

@github @code can you guys add o3 to github copilot plus tier as it is now cheaper than 4o , and like all of the other providers (cursor and windsurf made it x1 credit)

[random dude @nedjmi_]

@OfficialLoganK seeing the '> 200K' output token tier. Can this version of Gemini 2.5 Pro actually output more than 200k tokens, or does this pricing apply differently? Confused since I thought the max was ~66k.

[random dude @nedjmi_]

@alexalbert__ What about possible hallucinated citations? Or is it handled

[random dude @nedjmi_]

@_avichawla Very helpful, thanks !
Do you know of any open source projects for agentic Rag?

[random dude @nedjmi_]

@AravSrinivas Then you're a malware wrraper

[random dude @nedjmi_]

@btibor91 Can you use prompt cashing + batch api to get the 50% off twice 😂

[random dude @nedjmi_]

@OfficialLoganK i cant find the api docs for REST api, did you guys remove it ?

[random dude @nedjmi_]

Hey @alexalbert__, love your work! Just one thing: when there are network issues, the prompt and response disappear. It's really inconvenient, especially after spending time on a prompt. Can you keep the convo and just show a network error like OpenAI does? Thanks!

[random dude @nedjmi_]

@VikParuchuri Great system, second only to google vision which is pretty cool

[random dude @nedjmi_]

@yacineMTB Why's that? because that is so against the mainstream narrative "you either build your own dream or somebody else's"

[random dude @nedjmi_]

@phill__1 @ns123abc I feel like google definitely publishes more science, anthropic only publishes safety and mechanistic interpretability stuff i believe

[random dude @nedjmi_]

@ylecun People obsessing with end-of-the-world scenarios is a tale as old of time

[random dude @nedjmi_]

@StevenMusielski @ylecun @vkhosla He invested millions of $$ in openAi
https://t.co/f73r8irgBN

[random dude @nedjmi_]

@yumidiot Yeah sure, they have agi but they still can't have a stable product that doesn't throw errors at every other prompt..

[random dude @nedjmi_]

@OfficialLoganK Can you guys add system message in the rest api

[random dude @nedjmi_]

@zukkkk5 @Rousso03 @futuristflower Yeah i agree that most probably we wont see a leap like the 3 to 4 anytime soon especially with just auto regressive models

[random dude @nedjmi_]

@zukkkk5 @Rousso03 @futuristflower It's not only about the amount of data, there are a lot of other factors quality of data, more layers, more training runs, new model architectures....