Just came across this video showing an improved design for making strong boxes from laser cut plywood. I thought it was very interesting, and he supplies a Fusion 360 model in the links.
Found it while leafing through the Just Add Sharks blog, a UK laser company that recently closed. Their blog has many years of links to interesting laser cutter projects, though.
As youāre probably aware, one of the creators of Just Add Sharks is a member of the forum.
And he released some great stuff to the forum, which you may have missed:
He makes some good points and does a fine job of discussing joinery. Donāt know about the ācauterizingā thing. Might have to get Mathias Wandel to do some independent testing.
He also didnāt discuss the strength of finger joints when using solid wood rather than plywood. These hold up better than plywood even laser cut.
It is a good point to make. For boxes that will have a lot of diagonal stresses or twist, the typical finger joint isnāt going to deal with those stess completely.
Would be interesting to test a box made with two layers of finger joints, but with fingers offset in the layers to overlap differently. Of course you are going to have a stronger box with another layer over it. Come up with a design that uses one layer that is just as strong. Now that would be some cool engineering.
Oh yes, and donāt 
in my sandbox! Iām playing in it and I kind of like it this way.
I was not aware! I did see @m_raynsfordās other post, and leafing through it, noticed that many of the designs that had caught my eye from the forum had all been by the same guy. Itās an impressive body of work. And I came across the sharks blog completely independently today and had no idea.
Very true that any design that doubles up the walls is going to be stronger than a plain box no matter what joints you use.
Yeah, I really wish this guy could get his order. He would be magnificent with it.
That it is and heās a really nice guy. I met him last year at NY MakerFaire.
This was my main problem with the box. It used twice as much material to be twice as strong. Should have been comparing it to a 6mm single wall box
I suspect Iām way down the shipping list. I ordered my laser based on the strength of continuous autofocus and the ability to flip the material over to cut through 12mm from the other side.
I havenāt forgotten about these āimpossibleā features and that GF still havenāt even demonstrated itās possible in principle, let alone implemented it for everyone to use.
Iām also definitely not a GF cheerleader, Iāve seen enough lasers to be objective about my reviews and I get the impression thatās not the kind of feedback theyāre looking for at the moment. 
The feature I would have like to compensate for a lower power basic. I hope it goes live sometime.
I donāt even know how that would work with cameras and no fiducials in the bed. The more I think of it, the less I imagine that it will come to fruition. If I could only understand the defined scope and use cases that are intended by double sided cutting, maybe.
Itās like the pass through slot. At the moment in the interface there is nothing to indicate to me that there is any type of automatic placement of a design for materials moving through the slot. How would that even work?
Is it really possible for the lid camera and software to map the material and then assist in placing the design accurately over that map? I guess it is possible, but I donāt see indications of the software getting built out to it, other than some improvements in improved image processing.
Without those features the head camera is completely pointless. The material barcode can be read by the lid camera and both cameras are little more than a gimmick anyway.
Fiducial marks for alignment would be really useful for alignment. While I believe those things are possible (former life in vision recognition software) I notice that theyāve been secretly dropped off the list of functions. I bought the machine based on the strength of things they donāt appear to be supplying anymore. Itās almost pointless asking them if they intend to supply it too, without a proof of concept (which they donāt have time to do) they donāt know if itās possible to supply the features they promised to supply.
Oh I wish we had Fiducial recognition. You could even do just edge finding, based on the intersecting lines create a quadrant map of the work space.
What software did you use? At my work we mainly use Insight and Vision builder. Unless the machine came with itās own.
Weāre talking 1999-2005, specialised capture cards to get data into the machine at 60FPS, we were writing our own software to process images, do pattern matching etc. Open CV makes everything a lot easier now and you can squeeze it all onto a raspberry pi which is actually faster than the first machine we had.
In fact at the 1999 Robot football world cup in Dubai by coincidence everyone had identical machines, 400mHz CPU with 64mb SDRAM. It was definitely a squeeze doing perspective corrections and identifying 13 objects within the field of view at 60FPS.
The head camera measures the material thickness currently I believe. Does it even visit the barcode?
I donāt have a GF but I would guess that the machine gets the material thickness from the barcode,
What happens for focusing when you put in custom material?
I think it gets the material thickness from the barcode to make the camera view more accurate for placement. If you use non PG you have to enter it yourself.
When you render the design it says scanning. I think that is when it measures the actual thickness with the head cam and the red laser and uses that for focus.
People with machines please correct me if it this is wrong.
The camera on the lid is what scans the datamatrix. The information contained in that code tells the app that you have a proofgrade sheet, then populates the preset settings.
So the camera in the head is just a world of potential waiting to happen and currently serving no use?
I havenāt seen it do anything yet. When you hit āprintā it does āscanā your material. But that seems to be just a superfluous action. Maybe it is actually scanning, then feeding that information to Glowforge HQ and is being reviewed for future upgrades? Like Implementation of flip to etch. Donāt know. Just half hazarding a guess.
I am pretty sure it measures the thickness at the middle of your print and sets the focus.