Mohit Mor

Is Google still taking 5 rounds of DSA interviews? If yes, why? It’s strange how they’re leading the world in AI research and building some of the smartest models ever created… yet still filtering engineers based on LeetCode speedruns.

The token cost doesn’t seem evident in few tries but if you keep on using it continuously which you would since it is a mouse pointer, the cost would accumulate over days. When openclaw got famous people used agents to monitor flight prices etc for example and after few days realized just simple periodic monitoring costed them so much. I am not denying that there could be efficient and better ways to do things, i am just pointing out if these features that are being made, is cost kept in mind or not. But in any case this is really cool idea and i liked it too.

@mohitmor_ai @GoogleDeepMind Why would that need to cost a lot of tokens? or need a big model?. The pointer only sends what it sees after the trigger word, and they can use a small flash model. I like it.

We’re reimagining a 50-year-old interface - the mouse pointer - with AI. 🖱️ These experimental demos show how people can intuitively direct Gemini on their screens using motion, speech, and natural shorthand to get things done 🧵

@duncan_pkvk9 @GoogleDeepMind Local models are not that fast, and if we go with very small sized ones they probably are not that smart. I mean in future sure, but in short term it looks difficult with local models if we dont have a decent gpu.

@mohitmor_ai @GoogleDeepMind What if this runs locally

@kvickart @GoogleDeepMind True i think i was thinking more in terms of how feasible it is to adopt features for everyone in short term. And I was personally focusing on projects to save some cost so probably some of that bias :)

@mohitmor_ai @GoogleDeepMind my personal reference point is when i was a kid and I would be THRILLED to see an internet transfer speed hit 100kb/s, it was like a rare event and just look at where we are today

@kvickart @GoogleDeepMind I hope so, but with every latest model release the api cost is increasing, and all these features or future ones are good to use with smart models only. But yeah agree that overtime we will get smarter models like gemma4 which we can run locally.

@mohitmor_ai @GoogleDeepMind the cost will decrease over time, what used to be extremely expensive compute wise is barely notable these days

Google is starting to cook again but this is more like a cool feature. Probably in a week or 2 we will also see 3.2!

@OpenAIDevs More like a jira board for agents? Man i had made this long back for my custom agent github.com/mohit17mor/Arc I really need to up my marketing game 🥲🥲

Symphony: every open task gets a running Codex agent

@thsottiaux /pet feature was a copy of claude but? :p

The master becomes the mentee. At last, Claude is now copying Codex. But you cannot out-accelerate GPT-5.5.

@thsottiaux Maybe slowly start removing the need to manually surf the internet? Basically codex should be able to for example find a product, flight etc and give me all the information in the app itself in a nice dynamic ui based on the information

Now that the Codex app is close to being the super app. What should the super duper app do?

@karpathy How about dynamic UI based on information it needs to show? Html output it probably first step towards this?

This works really well btw, at the end of your query ask your LLM to "structure your response as HTML", then view the generated file in your browser. I've also had some success asking the LLM to present its output as slideshows, etc. More generally, imo audio is the human-preferred input to AIs but vision (images/animations/video) is the preferred output from them. Around a ~third of our brains are a massively parallel processor dedicated to vision, it is the 10-lane superhighway of information into brain. As AI improves, I think we'll see a progression that takes advantage: 1) raw text (hard/effortful to read) 2) markdown (bold, italic, headings, tables, a bit easier on the eyes) <-- current default 3) HTML (still procedural with underlying code, but a lot more flexibility on the graphics, layout, even interactivity) <-- early but forming new good default ...4,5,6,... n) interactive neural videos/simulations Imo the extrapolation (though the technology doesn't exist just yet) ends in some kind of interactive videos generated directly by a diffusion neural net. Many open questions as to how exact/procedural "Software 1.0" artifacts (e.g. interactive simulations) may be woven together with neural artifacts (diffusion grids), but generally something in the direction of the recently viral x.com/zan2434/status… There are also improvements necessary and pending at the input. Audio nor text nor video alone are not enough, e.g. I feel a need to point/gesture to things on the screen, similar to all the things you would do with a person physically next to you and your computer screen. TLDR The input/output mind meld between humans and AIs is ongoing and there is a lot of work to do and significant progress to be made, way before jumping all the way into neuralink-esque BCIs and all that. For what's worth exploring at the current stage, hot tip try ask for HTML.

Guide: @mohit17mor/how-to-run-ollama-on-free-cloud-gpu-guide-deae6287d97a" target="_blank" rel="nofollow noopener">medium.com/@mohit17mor/ho…

I put together a full guide covering: • Setting up Ollama on Modal • Using cloud GPUs • Exposing the API • Running models remotely This makes experimenting with larger local models way easier.

I figured out how to run Ollama on powerful cloud GPUs for free using Modal. No expensive local GPU setup. No managing servers. Just deploy and run open-source LLMs remotely. And yes you will have access to powerful gpus like A100, H100, etc so you will literally be able to run bigger models of size like 60-80GB as well!

@iruletheworldmo Gemini omni probably

it’s been way to quiet lately, i think it’s time for some huge releases this week 👀

Looks like gemini is going to release its new model soon!

@thsottiaux India when 😬😬

We are Codex. We are legion. We do not miss. We do ship. Expect us. DM me if you have a track record of solving near impossible problems and want to work on the future of computing. We work across research, infra, agent harness and products. Currently hiring in San Francisco only.

NEW: Australia to quarantine & repatriate passengers from the Hantavirus-hit cruise ship.

I always feel stuck at distribution, no matter what i build i am never able to market it properly, thats why i started posting on social media but now it feels like i am late in that too as i dont have followers or reach. In the AI world things are moving so fast that if you dont market your product, the idea might get implemented by someone else within few weeks. Any suggestions for me? 🥲

x introduced me to building businesses when i was 16. built and sold two businesses, which enabled me to come to the US. then it got me a job at @cloudflare. and I got my first 100k users on my own little sideproject. which later became @supermemory. and i met most of my investors through X as well.

Too long since gemini upgraded 3.1