Aditya Sohoni

@TheSalonDon @grok where is the essay?

This guy is on NZT - Valedictorian of Columbia University at 19 years old - Joins OpenAI on 8 figure contract -Raises concern about data security -Fired -Writes 165 page essay about next 10 years -Raises $200m to invest in the strategy -Turns it into $5.5b in 18 months Leopold Aschenbrenner 24 years old Remember the name

@richzou amazing, thanks for the effort!

would highly recommend taking a read kokospotlights.me/wp-content/upl… kokospotlights.me/wp-content/upl…

@eurofounder there is a massive opportunity here

My daughter got an offer from Google for a summer internship in California "They will pay me $12,000 per month" she was so excited While the whole family was celebrating, I excused myself I locked myself in a toilet and called Google's HR "My daughter received an internship offer from you, but there must be a mistake" I said The HR woman insisted this salary was standard for engineering interns "No company pays a young woman that much without expecting something in return. Do you want to prostitute my daughter?" I raised my voice She went quiet "Who are you again?" she asked I told her to withdraw the offer immediately My daughter found out three days later and hasn't spoken to me since I already sent her CV to Siemens and SAP, respectable technology companies, much bigger than Google She'll thank me when she's older

@mukundjha @emergentlabs @khoslaventures @SoftBank amazing, congrats!

15 months ago I didn't expect I'd start another company so soon Then I saw how many great ideas never make it past day one Today, we’re announcing a $70M Series B for @emergentlabs from @khoslaventures & @SoftBank Here's the story of how we got here (thread) ⬇️

@ruomingpang @mattdeitke congrats both of you!

@mattdeitke Looking forward to working with you, Matt!

Last week, I joined Meta Superintelligence Lab (MSL)! This is going to be an exciting ride! 🚀

@garrytan very well put

There’s more to life than money, but it takes money to find out what that is.

@TechEmails fascinating thread. shows you zuck is really good with understanding money.

[This document is from FTC v. Meta (2025).]

@gsivulka really? which ones?

In case you missed it: Every big AI lab is now training domain-specific models... Telling sign that we've reached a wall scaling pre-training.

@aryxnsharma this is the real business opportunity

looking for 2 typescript backend engineers who have experience in building third party integrations. ideally can start immediately. open to remote. please dm!

@levie steve jobs was so simple yet so brilliant

Steve Jobs explains that every new medium initially copies the last medium, but then we figure out what it’s much better at. This happened from radio to TV, mainframe to PC, and PC to mobile. The same will happen in AI as it changes how we use software.

@elon_docs this would not be possible today

Elon Musk: The only reason I started a company was that I couldn't get a job anywhere. “The only reason I started an Internet company back in 1995 was because there were only a few Internet companies and I couldn't get a job at any of them.  I tried to get a job at Netscape and sent my resume in and tried hanging out in the lobby. But I was too shy to talk to anyone.  Then I was like, okay, I guess I'll have to start a company because I can't get a job anywhere.” Montana Jobs Summit, September 16, 2013

@tsarnick damn, stock market crash incoming?

Marc Andreessen and Ben Horowitz say that AI models are hitting a ceiling of capabilities: "we've really slowed down in terms of the amount of improvement... we're increasing GPUs, but we're not getting the intelligence improvements, at all"

@chris_j_paxton nice

Another cool example of how modern large language models make stuff that used to be years' worth of work possible for a tiny team in just a couple days This will power the coming explosion in robotics

currently going the hard way and googling each citation and concept....

startup idea: create an ai that can give beautiful explanations of such papers. should also explain all the background math required to even start reading something like this

tried reading arxiv.org/abs/2210.02747 wtf is this paper

@c_valenzuelab yeah turning something using your hands is really intuitive for humans also is making objects in the air and pointing to them while explaining something would be very interesting to see how ai can capture all that

Whatever prompt strategy seems cutting-edge today will likely be irrelevant tomorrow. We're discovering entirely new ways of generating. A prompt is just one possible compass, and perhaps not even the most useful one. Think of any large model as containing an intrinsic infinite world - a multidimensional space with countless pathways and connections. The real challenge isn't crafting the perfect prompt – it's developing whatever vehicle is needed to effectively navigate this space. Some of our most profound forms of expression are non-verbal. Why should our interaction with AI be limited to words? The future of AI interfaces won't be about writing better prompts. It will be about developing richer, more intuitive ways to navigate these vast latent spaces. Tools that understand gestures, that respond to visual feedback, that can engage in a dialogue of showing rather than telling. Intuitive exploration rather than a verbal guessing game. Act-One and Advanced Camera Controls are steps into that direction. Think of Advanced Camera Controls as finger frames: directors' hands communicating more intuitive ways of expressing your intentions.

@c_valenzuelab how will you ever handle competition from those who have large gpu clusters and create massive image and video models like meta movie gen? you guys sure must've discussed it, really curious to know your thoughts

Runway is not an AI company. Runway is a media and entertainment company. And I actually think the era of AI companies is over. It's not because AI failed - quite the opposite. It's because AI is becoming infrastructure, as fundamental as electricity or the internet. Calling yourself an AI company today is like calling yourself an internet company. It's meaningless because it's universal. Every company uses the internet; every company will use AI. For Runway, our focus is art, media, and entertainment at large. We began Runway almost seven years ago with a vision that remains largely unchanged today: AI is a necessary tool for storytelling. To achieve that vision, we had to work backwards to build the best research team that could deliver the best models on which we could build the best products. I often talk about our work as a new kind of camera. Not in the literal sense of capturing images, but in terms of its historical impact. The camera didn't just create photography - it birthed entire industries, economies, and art forms. Cinema, television, TikTok - all children of that first revolutionary tool that could capture light and time. I think the work we are doing at Runway is part of a new foundation for an entirely new media landscape. Just as the camera transformed how we capture reality, AI is transforming how we create it. The models and technical capabilities we've built are just the beginning - they're the equivalent of those first daguerreotypes, primitive yet pregnant with possibility. The mistake many make is seeing AI as the end goal. It's not. AI is the mechanism, the underlying infrastructure that enables something greater. The real revolution isn't in the technology itself but in what it enables: new forms of expression, new ways of telling stories, new methods of connecting human experiences. Media has traditionally operated like a one-way street. Creation flows down established channels to reach consumers. Even when distribution was disrupted - first by social media, then by streaming - the fundamental pattern remained: someone creates, others consume. The roles were clear, the boundaries defined. But we're witnessing something different now. Imagine watching something that generates itself as you watch it - truly dynamic content that responds to you, understands you, creates for you. Universal Simulation and world building. The distinction between creation and distribution dissolves when content can shape itself in real-time. That is the foundation for an entirely new media landscape. It's about fundamentally reimagining what media can be: interactive, generative, personal - yet simultaneously shared and universal. This is also why pure AI companies are becoming obsolete. The interesting questions aren't about the technology anymore - they're about what we build with it. The next wave of innovation won't come from companies focused on building better models. Models are commodities. The technical foundations are now well-established and known by everyone. There are no secrets. The wave of change will come from those who understand how to use these tools to create new forms of media, new types of experiences, new ways of telling stories. The infrastructure is laid. The foundation is built. Now comes the exciting part: creating something meaningful with it. The end of AI companies marks the beginning of something far more interesting: the birth of truly new media. Not just new platforms or formats, but entirely new ways of creating and experiencing content. We're not building an AI company. And that's a far more exciting mission. Like it has always been; back to our roots.

sometimes I feel like guys like @alanmnichol actually make sense, that intent-based chatbots are better than llm agents. But then I'm like nah llm agents are the future. but one thing is clear. llm agents today are not business ready