The magnitude of this precision under the force of gravity for such a monstrous object is mind-blowing. I was reading to see who should be attributed for this challenging idea, and it seems it really was Elon, with support from Stephen Harlow. A lot of the time, I think, "Elon, the greatest entrepreneur of our lifetime," and leave it at that, but I think that’s actually shortchanging the reality. It really should be: "Elon, the greatest entrepreneur of our lifetime... and engineer who made the unthinkable thinkable."

@damengchen What are they using in lieu of cursor? I l'd like to bandwagon that.

@fekdaoui Cool, but it's not free... I still need to pay for API calls.

Eeehh, not a good idea to contract out a commercial that doesn't use your own product that is suppose to make such video generation feasible. In other words, shouldn't ChatGPT+SORA have built this, lol?

@tunguz "make a chip ... easy part". Nah, this is hard, well in the sense you have to do it better than Nvidia's Hardware+NCCL+CUDA

@TimothyDuignan Haven't actually done this yet, how well does Orb capture the melting temperature transition?

@MarinaMedvin Why not just ask Chile, they have one of the largest diasporas of Palestinians, something like ~500K.

@BenjaminDEKR I think for small companies that can take the security risk, building your own SaaS as you need it, will work and makes sense, but will the 1000+ employee company take such risk? Or I guess the future of companies is 50-100 people doing it all.

@sama Interesting, just hard to imagine observations hold, in a practical sense, w.r.t physical infrastructure as it can't scale indefinitely. Also seems a new paradigm of compute hardware will be needed.

@BenjaminDEKR Yes I am.. TL;DR Don't teach my kids to code, teach them to use AI tools that aid them in how to create and manifest their ideas... And along they way they pick up some coding basics.

@ryolu_ @cursor_ai @shaoruu So sonnet 3.5 or o3-mini

Scientist/researcher probably isn't as safe as one might think, especially at junior level. I think a lab technician is more inline with a job o3 can't replace. But once autonomous robotic systems become ubiquitous in the lab, science will be done/lead exclusively by theorist, AI systems will do all the experiments/computations/data analysis.

Don't work in hep-th but I've been told in the past by senior people who I've done research with that in my presentations I tend to be pedantic, i.e., I cover a lot of nuance & detail so that researchers can indeed follow along, even if they aren't working in the topic. They basically were telling me I was talking down to audience, but I just thought I was being detailed & thorough. It's one of the reasons I hate verbal presentations and rather just write things up.

This must be founder mode, one doing something that test your abilities and another saying don't quite !

@rohanpaul_ai I want to study the sub bullets, what would be the list of books and/courses to learn about them

@AverageProMax @cursor_ai @OfficialLoganK @GoogleDeepMind I just don't see theses models working well in compose mode. Half the time they don't propose changes to apply and they code lazy

@krishnanrohit It's too verbose though, it never gives me a concise direct answer it always wants to blabber on.

@BenjaminDEKR Have you used o1/o3-mini in Cursor? It just doesnt' seem that good compared to Sonnet, your thoughts?

Creating an app that at 4:30 pm on Friday, every week, flashes this picture for 30 seconds that also uses @elonmusk voice to say: "What did you get done this week?"

Very nice opinion piece, but my take differs. Universities exist primarily to advance knowledge and serve the public good. Whether they make or lose money is tangential to that mission. Their funding, through endowments or state support, reflects this. Profit-driven entities like Bell Labs assumed all commercial risk of success on projects, hence 100% indirect costs. That is, if the DoD funds R&D on X and provides $Y and X fails, they have no way to make money from it. Full cost recovery offsets that risk. In contrast , universities conduct research as a public service; companies do it for economic gain. As for folding indirect costs into direct, I see no issue. Forecasting and resource management are standard practices to estimate costs—challenging but feasible and transparent.

Administrative bloat must be addressed. Over the past two decades, academia has steadily expanded its support staff, often without clear necessity, creating the illusion of improved efficiency rather than delivering substantive benefits. For instance, has grant writing become more streamlined or accessible due to the influx of staff dedicated to assisting with grants? I would argue barely. Similarly, while securing lab space or managing projects may have become marginally easier, the improvements are far from transformative. At public land-grant universities, the responsibility for covering indirect costs should rightfully shift to the state. Most private institutions have endowments that grow at rates that can cover such cost.