Elon gets the idea. https://x.com/elonmusk/status/...

Ummm... @elonmusk, what do you think about this idea for training one of these "super LLMs" I'm describing, that will be able to achieve this concept of "Any input bitstream to any output bitstream"? 👇 "Since LLMs and agents are close to generating end-to-end apps (backend and

"But banks still run on COBOL." Ok, let’s clarify: with "all software will run on the LLM layer". Common sense: I’m talking about 99% of software… except for super-specific stuff like code for nuclear power plants, banks, hardware controllers and things like that, which we

@taherdhanera Which one was yesterday?

@javilopen Genie3's researchers explaining that the temporal coherence wasn't the result of gaussian splat, a game engine, or clever context working but intrinsically learned is when I fully joined the "yep, the entire stack will be generated" crowd. So pass the slurpy and fez over...

@tariusdamon Amen, bro.

@javilopen Well I guess you must have loved the painting exemple of genie 3 with the persistent memory

@Hugal_ Exactly! Great example!

@javilopen Wild take but honestly feels inevitable at this point

@javilopen I love the energy Javi but the future is more likely to be LLMs calling software, not LLMs replacing software. Core computational engines, safety-critical systems, and high-performance workloads will still be implemented in specialized code, not run “inside” the LLM.

@javilopen I can totally see this! Having similar discussions with people in the media/content creation space.

@javilopen You are very optimistic tbh. We tend to oversimplify the steps necessary and oversell the outcomes. AGI was supposed to be around the corner, same for other achievements. As I’ve seen in 40 years since my first program on vic20, things will be way more complicated (and exciting)

@javilopen “The economy will dissolve into a planetary ASI layer that ends both money and human hierarchies.” What does that even mean? I’ll break it down in detail (long post incoming . . !) and once you finish, there won’t be any middle ground. You’ll either reject the idea completely

@javilopen Fascinating vision. Feels plausible, if not inevitable. I was toying with the idea that LLMs would just directly write machine code, but indeed, why even bother 🤷‍♂️

@javilopen If I had read your post before Genie 3 I probably would have laughed or called you crazy. Now I’m not so sure anymore, you might be right. Then again, looking at GPT and all the benchmaxed models, it will definitely take some time until we get there.

@javilopen One thing yet to be demonstrated is delivering software that does what you want it to do rather than what you tell it to do. Human engineers routinely solve problems they don't know how to solve. AI, not so much.

@javilopen It’ll basically be a persistent simulation of anything.

@javilopen Internet should become a commodity first

@javilopen I dont understand how some people can’t see this. Same happened with art, video, etc. they don’t learn the lesson

@javilopen Isn’t entertainment a form of art? Are we saying art gets commoditized too? I understand the premise but still trying to figure out how how art becomes a commodity

@javilopen Tienes toda la razón Javi, nosotros hemos creado https://civiges.albages.cat sin tocar a penas código.

@javilopen yeah bank app run over a LLM ........

@javilopen Thanks for 'cooking' the 'dish' and share it to the world @javilopen. It's so 'delicious'... 😊🙏

@javilopen Just take a look into @torus_network and @tigfoundation Thats enough to say ♓🕸️

@javilopen Makes a lot of sense. The UI of LLM will also change/evolve. Most humans don't do conscious thinking, which is required to us LLM eddectively, they are victims of interfaces, and what the interface shows. So LLM will perhaps have ability of shapeshifter as per the needs of user.

@javilopen And we need to think with a different mindset https://x.com/DrJimFan/status/...

@javilopen „l run directly in the LLM layer“ - absolutely, i malso convinced about that!

@javilopen Software will become to culture what memes are today. Instant, transferable, disposable. It'll require a massive shift in mental model but the change is inevitable, and that's sites we engage with. In parralell, agents will build invisible substrates we never see to reach goals

@javilopen So the LLM will generate everything or is it everything combined in 1? Either way you’re right

@javilopen The value of creative and intellectual output is steadily eroding in our modern world. Texts—whether meticulously crafted books, carefully translated works, or in-depth journalism—are increasingly undervalued as content floods digital platforms. Images, once the domain of skilled

@javilopen Makes a lot of sense. We have a long way to go, but with the compounding of LLMs designing newer LLMs exponentially it could be here way faster than we think.

@javilopen Agree. Anyone who truly knows how modern LLMs are built from scratch knows what you say could become reality soon. Those who dismiss it likely don’t understand the process, if they did, they’d see this technology can be applied to this goal and it would be just a matter of time.

@javilopen Let's hope at least calculator will still be calculators tho, cuz look at how gpt 5 can fails at simple math like 5.5-4.4

@javilopen Basically Jarvis with a few companies providing the infrastructure for what billions use with millions of jobs gone