Yes they definitely claim superpowers in their advert.
- Precision within a kerf
The laser works by removing a hairsbreadth of material, called the kerf. Glowforge knows this and compensates, so if you draw two puzzle pieces, they’ll fit. Even better, the Glowforge can measure the material thickness so 3D objects slide together perfectly.
But of course they haven’t rolled it out yet, so I would say it is a pipe dream.
Technically… they do have precision within a kerf. What they don’t have is automated ways to leverage it. It’s totally feasible to move he laser a thousandth of an inch or two with great precision. Kerfs are several times that in most materials.
To summarize, lasers are not straight 
Yes, If the material is acrylic and intended to be fused/glued, then a saw cut is best because of that slope. A saw cut gives flat edges - most surface area contact. Wood is of course the same, but with compensation (I do 0.007 on proofgrade ply) it is tight, and I like to use CA glue after fit-up.
But they use the word “compensate” in their marketing. That implies actively doing something, not just relying on things to almost fit without doing anything.
Printtolaser,
So sorry I am being a little thick here, but my machine isn’t here yet and as such I can’t even check out the UI online. When you say you use .007" to compensate and get good fits, are you preparing your drawings with the fingers oversize by .007, or is there a way to tell the UI to cut .007 outside the lines of your on-size drawing?
If the former, your drawing of, say, a 3/8" box joint would have a tab .375 + .007 = .382 wide. If the latter the. You would draw the joint .375 and adjust the path in software.
I plan to draw something up to test out when the machine gets here, so it would be helpful to know. Thanks.
Unless something has changed in the last week, that is up to you. Need to correct for kerf in your files.
Sadly I’ve run into a case where this is not true, either. I’m cutting some 1/4" exotic hardwood (Padauk), and the kerf seems to be different depending on length of the tabs and/or direction they run on the laser bed (horizontally or vertically). My initial testing showed a kerf of 0.2mm on some small pieces, but actual pieces I cut wouldn’t fit. I dropped the kerf to 0.1mm, and half of the joints worked and the other half still didn’t fit. I’ll probably have to reduce to 0.05mm or maybe even 0 and use glue to make it stick.
OK so:
There’s a lot of material on the forum about how to manually correct for kerfs using either path offsets or setting stroke width, converting stoke to path and removing the inner or outer paths that result as needed. I’ll try to find the links to those threads in a second but it’s as easy-ish as searching for “kerf” and just investing the time to read through it all. It is a bit daunting.
That being said, many box making sites have options to set kerf width. There are also lots of posts about box generators in the forum, again it’s just a matter of investing the time to read them all, which can be a lot.
So let’s see… kerf. Read this beast of a thread, you’ll be nearly expert by the time you explore it:
Box makers… I posted a roundup somewhere… hmm. Aha here we go, read this sucker:
And these are just one example of each. There are many other threads if you go digging.
Could that have been related to grain direction and hitting the soft layers be dark layers of the grain? I’m going to say yes, but it’s just a guess?
It’s certainly possible it is grain-related; I haven’t done exhaustive testing since I have a limited amount of wood to use 
I’ve also had a result like this with a box maker. At the time I figured the box maker had a bug and cut the tabs incorrectly on the bottom and top edges (way too loose). Maybe it was grain-related kerf discrepancies after all? IDK, hasn’t happened lately, and I don’t recall which generator it was. Bad scientist! Go to your room!
Sorry I wasn’t clear, I I was referring to the settings in inkscape tabbed box maker under extensions.
It’s not that kerf can’t be computed and designed for, it’s that there’s no magic answer that allows you not design for it. That’s true for 3D printing, laser cutting, routers, etc.
In my experience, parametric designs “win” at this, because you can make the needed slot width a parameter, and tweak it to suit the width of the material and the “tightness” that you want, and since different materials that are “the same thickness” still vary significantly, you need to actually try the material to see how it fits using the real material in the real world. This isn’t just a matter of cutting inside/on/outside the line, it’s that the width of the slot you’re cutting needs to match the actual material thickness, and achieve the degree of fit that you want.
Pretty much every time I have to cut a little tab and slot and fit them together, starting with measurements and then tweaking until the material “feels right”. More with wood, but even acrylic can vary between sheets that are “the same size”. If a loose fit is ok (e.g. you’re using glue) it’s less of an issue, but for a press fit, tolerances are quite subtle.
There’s theoretically a solution for Proofgrade material, in that if the material thickness is absolutely consistent, you could make a design that fits exactly for the material. And there were hints that you’d be able to make designs with marks for slots, which the GF software would make the right width to suit the material being cut. If the do that, it’d be nifty!
But until then, parametric designs, with small test pieces to test, is what works for me. On the Glowforge, on a router, and on 3D printers.
I’m using my own extension for OpenSCAD so I know this isn’t the software, at least
Couple of things…in traditional woodworking, the thickness variations of materials are usually accounted for by cutting your slots a little deep, and then sanding the tabs flush once the joint us assembled. That us no problem at all.
The finger width though, can be cut predictably, repeatedly, and you can be pretty darn sure they will go together.
What I have heard others say is that even if you cut the same material with the same setting two different times you can get two different dimensions because the laser kerf varies. That is difficult to manage. Or it seems so to someone who has not tried yet (me).
That is not something I face with 3d printing or a cnc router. Can chalk it up to part of the learnng curve with lasers, but that doesn’t mean I can’t wish for kerf compensation without having to modify my drawings. I use Alibre Design for my drawings, so it is easy enough to draw things the way I would for 3d printing or just general fabrication, and offset the drawing by a few thou to compensate the kerf, but it is still not as good a solution as other CAM software I have used that can be told to cut on the outside or inside of the line, with an offset of my choosing. Hopefully GF software team make good on their marketing some day and give us some kerf compensation in CAM.
May be I will get the machine and figure out it isn’t that big a deal after all.
Brian
Well it all comes down to your tolerances. We’re generally talking about a couple thousandths of an inch— with wood that’s basically the difference between a zero clearance glueup joint and a glue freee friction fit. I tend to shoot for friction fits which is always going to be tricky.
“Wood is alive”, as my woodworking buddy John says.
So yeah depending on your needs, kerf may not be an issue for you at all.
@bbasiaga: If you cut the same material with the same settings, you should have the same kerf. The kerf varies when you cut different materials, or use different settings, as you would expect.
You’re right - for cutting fingers, material thickness doesn’t affect whether the fingers ‘fit’. Though I’d point out that the length of the fingers should match the thickness of the material so that the sides are flat - you don’t want the “fingertips” sticking out!
That’s the whole point of sanding them flush after setting the joint.