Harsh Sharma

AI Product Manager

[Harsh Sharma @HarshHNH]

@FakePsyho So true man, people have been talking about the regulation of AI for so long,
No idea why it hit people's head only after demis's post

[Harsh Sharma @HarshHNH]

@demishassabis 100% there has to be a universal body to monitor and publish models, but how do we solve the open source risk? Considering future will enable local inference machines affordable @demishassabis ?

[Harsh Sharma @HarshHNH]

@joshwoodward @GeminiApp Ah many things man, have a solid plan for what you can do next to probably gain the market. Maybe an email which i can mail to pls?

[Harsh Sharma @HarshHNH]

@chetaslua On what grounds did you call Mythos hallucinations king?

[Harsh Sharma @HarshHNH]

@chetaslua @mafabiorx @ArtificialAnlys @chetaslua any comments on above comment?
Why are you lying about hallucinations?

[Aiden Bai @aidenybai]

just so we're clear:

Antigravity is a Windsurf wrapper

Windsurf is a VSCode wrapper

VSCode is an Electron wrapper

Electron is a Chromium wrapper

Chromium is a C++ wrapper

C++ is a C wrapper

C is an Assembly wrapper

Assembly is a Machine Code wrapper

Machine Code is a Binary wrapper

Binary is a Physics wrapper

Physics is a Math wrapper

Math is a Logic wrapper

Logic is a Philosophy wrapper

Philosophy is a Humans wrapper

Humans are a Carbon wrapper

Carbon is a Star-forged-matter wrapper

Stars are a Gravity wrapper

Gravity is… definitely not an Antigravity wrapper

[Harsh Sharma @HarshHNH]

@AravSrinivas @AravSrinivas How is factual accuracy benchmark calculated? Isn't it a bit of a red flag that it's too low for all models

[Harsh Sharma @HarshHNH]

@koltregaskes Thanks for the update! 😃

[Harsh Sharma @HarshHNH]

@OfficialLoganK Meeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee

[Harsh Sharma @HarshHNH]

@kernelKain @viditchess @AravSrinivas How is comet not a task agent?
It does tasks which atlas can't do right now.

Think more before commenting

Also,
@viditchess try using these browsers less for tasks right now, as they dont have guard rails for prompt injections built that well right now

[Harsh Sharma @HarshHNH]

@sama Request to talk more deeply and less high level. It gets annoying nowadays to hear same things repeated again & again in podcasts

[Harsh Sharma @HarshHNH]

@kevinweil I feel a simple feature like,
Setting AI personalised name in configuration,
And using that name as a trigger word when AI voice mode is on will be a great step in right direction

I can say " x" what do you think about it,
And it having context of whatever happened can jump in

[Harsh Sharma @HarshHNH]

@kevinweil Keeping above constants in mind,

I tried using voice assistant on 24/7 to see how the companion use case feels right now,

Architecture allows all inputs giving an output,
So voice was jumping in conversations when it wasnt meant to

[Harsh Sharma @HarshHNH]

@demishassabis @GeminiApp @OfficialLoganK @koltregaskes It will be a step in the right direction, as it will know what the conversation is about and can jump in whenever asked to

[Harsh Sharma @HarshHNH]

@demishassabis @GeminiApp Hey @demishassabis @OfficialLoganK @koltregaskes was trying to use gemini as a constant ai companion in life, but as the architecture allows replies to all inputs, it also replies when im not talking to it,

Can we like have a trigger word which we can set as configuration?

[Harsh Sharma @HarshHNH]

@Prashant_1722 @tetsuoai lol kid spotted.
I'm only talking about image modelling as the tweet is only for image.
Think more before commenting

What is your age? 13?

[Harsh Sharma @HarshHNH]

@Prashant_1722 Dude its no where real looking, nano banana is working way better. Even other models like Qwen, Kontext are better.
Stop over hyping every tweet

[Harsh Sharma @HarshHNH]

@icreatelife P e a c e