> After almost two decades of searching, a team of physicists in China says it has observed strong evidence of a mysterious particle called a glueball, which is made entirely of force-carrying particles.
> ... The discovery of glueballs would provide direct evidence that gluons can interact with themselves, researchers say. This is a key prediction of quantum chromodynamics
It’s the theory of colour charge essentially, counterpart to electrodynamics. Quarks have colour, rather than electrical charge. Mediating particle is the gluon, analogous to (but entirely different from) the photon
Its called chromodynamics because the interactions are described by the special unitary group of degree 3, which is the same group that describes how we perceive color interactions.
It's just the three dimensional nature of it plus the usual physics practice of giving things cheeky names. The vector space of (visible) colors is not acted symmetrically upon by SU(3); it doesn't transform symmetrically under charge-preserving unitary matrices. A blue strawberry with red leaves would not seem the same to us as a red strawberry with green leaves.
1. A frontier model wrapper that gens the designs
2. A bunch of humans behind the scenes ("AI means Another Indian") actually reviewing and fixing the design.
That is, get to market quickly with people, gather lots of data, and hope that models can be refined eventually so less/no human review is needed.
I recently worked with Claude Opus 5 High to help me build a testing rig for some acoustic tests. There was a lot of back and forth with each time Claude giving me the stereotypical "Great catch! I see my mistake in the previous iteration..." responses.
How could a person (of any ethnicity) be able to respond reliably enough and quickly enough to design nudges to be able to iterate back and forth on this kind of thing with any usable latency? I don't think it's possible.
working on it, raising our Series Seed now at xfgr.ai
a proper foundation model will operate purely in the latent space to output a STEP file. there will be no LLM written cadquery/openscad.
there is no other way to scale, give us 5-10 years.
I mean, there's marketing copy on this page that makes me think the team doesn't really have any real manufacturing or design experience in the real world.
>>"Make this enclosure lighter while keeping it strong.
Optimized the geometry while maintaining structural requirements.
31% lighter
Structure verified"
Are we supposed to believe the AI just intuitively knows how strong is strong enough for the hypothetical enclosure? Engineering is about designing against actual constraints, and often those constraints are the actual hard part. For example, determining how strong an enclosure is is very easy, analysis tools in CAD/Cam software have had that for years and calculating yield is a 1st or 2nd year undergrad problem. The actual hard part is knowing how strong is strong enough. Rarely, if ever, are actual limits given in hard units. The engineer needs to determine how strong the enclosure needs to be to survive the intended use, and balance that with cost. But those design limits normally aren't given as "this needs to withstand this many kPa of point load", determining what that load is is often the hard part. There seems to be a lot of hand waving going on with this marketing copy and absolutely no details, only a 'book a demo' link.
I'd also be very, very skeptical of how useful and modifiable the design files produced are. It's really easy to crap out a design in parametric CAD that is un-modifiable. Sketches can be done poorly or the model can be structured badly, so it becomes hard to edit later. I notice that their example files are all 3d meshes and not proper design files. If I'm making a physical product, having just a STL file doesn't help me much if I need to actually modify 'my' design. Am I to believe they can actually export a Solidworks file? I'm skeptical.
While in theory it's a cool idea, this marketing copy amakes me think this is a vibe-coded project by a team with very little real-world experience. They are likely going to struggle landing any real customers, since their sales team won't be able to answer the questions a serious customer is going to ask. However, they will probably book some demos from people who have a idea of a product but no real road map. As someone who has done software consulting adjacent to this space (focusing on small manufacturers and design houses that would be their customers), the type of customers that will book a demo off of this page are also the type of customers you don't want to have.
an agent could run a finite element analysis or some kind of generative design algorithm, it's not new stuff all of it is already available in Fusion by Autodesk.
> Generated text will carry embedded watermarks, and generated files will include digitally signed provenance metadata where supported.
This should make it easier to catch cheaters who use Claude, right? Unless everyone runs their artifacts through some watermark and metadata sanitizer?
The approach Pangram has taken which works pretty well is to simply lower the recall a lot but ensure the precision is very high. Which means potentially high false negative rate but low false positive rate.
> An unreleased research version of Claude has improved on a longstanding lower bound for the fraction of zeros of the Riemann zeta function that satisfy the Riemann hypothesis. Drawing on extensive prior research by mathematicians over the past decades, it has increased this bound from 41.6% to 67.2%.
If I recall correctly, at an undistorted scale, the water would be so shallow that surface tension and viscosity would dominate, so the depths are exaggerated to keep the flow realistic.
More specifically, tidal flow obeys Froude similarity, not Reynolds. Matching the model's Froude number to the real Bay's requires enough depth for gravity waves and tides to scale correctly, which the vertical exaggeration provides.
But the distortion makes the flow too efficient, so copper strips are added throughout to achieve the right frictional resistance.
reply