Sabitlenmiş Tweet
*tess
6.2K posts
*tess
@ptr
alea iacta est. my views are my own, don't take them they're all i have left!
🦋 Katılım Haziran 2008
920 Takip Edilen5.1K Takipçiler
@ptr Super simulations:
x.com/doodlestein/st…
Jeffrey Emanuel@doodlestein
This project is just getting more and more wildly ambitious. At this rate, it will soon be a full-fledged, agent-first replacement for SolidWorks or Autodesk Fusion, but with way more knowledge of physics and a much more advanced architecture based on the math of sheaf cohomology. And more importantly, designed from the ground up to be used by agents instead of by human users in a GUI. And instead of using the system to conceptualize, implement, and simulate a single design, a swarm of agents could explore a universe of designs. My dream with this now is that I can use it to shepherd the design of a radically better electric motor or generator or other useful contraption using completely new ideas and designs, and then test it in the real world (anyone have a line on a good metal 3D printer?). Imagine a team of 20 Sol Ultra and 20 Fable Max agents working together for a few weeks with a bunch of fast computers to "find" a better design in the vast computational landscape of possible designs, guided by an optimization pipeline that is only possible because of the unification of the various simulation kernels into a coherent system.
Français
@_lopopolo if we invented ai first and software second, everything about how software is managed and deployed would be different
it's rife with anachronisms
English
The overhang is large and only growing. Deployment remains (one of) the hard part!
Logan Kilpatrick@OfficialLoganK
every ~3 months, you need to increase your level of ambition in the AI era, else you forfeit the capability overhang of the models to your competitors
English
@ptr What was most exciting consumer product experience in the before times
English
@thenathancolo no no you don't get it, i'm asking *you* to be bored with small problems and solve bigger ones
English
@ptr If it’s solved then why do I spend all day reading code that doesn’t work
Just solve bigger problems if small ones are solved, ofc continuing to solve small problems will be boring
English
@ptr ai + science. drug discovery, materials science, bio engineering.
ai + creative tools. not prompts but actual tools to help artists, musicians, etc make beautiful things.
infinity stuff to do to improve the human condition outside saas to make more saas.
English
tech companies often say they want to 'democratize access' which means 'we want a lot of users' as a sort of proxy metric for world conquest. FB cared about DAU more than revenue too. not quite a standin for virtue
should use the earth's limited computational resources well
English
@vijayiyengar so true
i really am hoping we will see a pivot to harder things as ai saturates
the constant multiplexing / context switching is a big barrier to that … if frontier ability was blazing fast it would be easier to marshal grey matter on hard things
English
I think lower latency via better hardware and voice models will make all sorts of new experiences.
Fable intelligence at sub-second iteration speed is likely worth a lot:
- Less context switching
- You'd be more in-the-loop which would yield better results
- You could test explore more ideas thoroughly and rapidly
At that point, I think we'd actually be creating 10x more useful software per unit time and need a rethink of all the surrounding systems (CI, analytics, etc)
English
@tokenbender mmmm. it's a second derivative compression huh
floor of interestingness rising and pool of people who can generate ideas in the remaining space shrinking
English
@ptr it is hard to be impressed by a problem of choice now, what captivated one's attention one year back is a /goal now.
so our own standards are only stimulated by something fresh and creative now.
English
@tokenbender shrinking like, no more ideas
or shrinking like, no more people working on them
English
@ptr building interesting things is a rapidly shrinking market. but we’re primarily bottlenecked by creativity for the time being if one is outside of frontier labs.
English
UI Components for web that replicate terminal ui
Română