Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
#437 White light interferometer and Grimsmo Vision
2.1k Plays2 days ago
Topics:
- White light interferometer and Grimsmo Vision
- Solving other people's problems, not just your own
- Saunders travelled to Lights Out Automation, Metalquest, Autodesk, Lambcrafted, and Johnny Five
- New Speedio and UMC
- Historical financials
Recommended
Transcript
Introduction and Business Updates
00:00:00
johngrimsmo
Good morning and welcome to the business of machining episode 437. My name is John Grimsmoke.
00:00:07
John S
My name is John Saunders.
00:00:08
johngrimsmo
And this is your weekly dose of manufacturing and little bit of business sprinkled in there um about our manufacturing businesses and the journey we're on and what we're up to and what we're thinking about and what we're working on.
00:00:21
John S
Yes.
Returning from a Break
00:00:23
John S
And we're coming back from an unusually long break. I feel like we've we've taken one week off, but yeah, a couple weeks is more than normal.
00:00:25
johngrimsmo
wow couple weeks off.
00:00:29
johngrimsmo
ye Yeah. That's good though.
00:00:30
John S
How you been?
00:00:31
johngrimsmo
We're allowed.
00:00:32
John S
Yes. How you been?
00:00:34
johngrimsmo
I've been good. Been good. Things are steady. Things are
New Projects and Innovations
00:00:37
johngrimsmo
good. Been playing with a lot of really interesting projects that I can't wait to dig in and get you excited about.
00:00:41
John S
Okay.
00:00:43
johngrimsmo
um
00:00:45
John S
Do tell.
00:00:45
johngrimsmo
Yeah. i have lots I have lots to update you.
00:00:48
John S
Okay.
00:00:49
johngrimsmo
um I'll give you the the list first. um White light interferometer, Grim's MoVision.
00:00:54
John S
e
00:00:57
johngrimsmo
Cursor as a programming tool and visiting Hacksmith yesterday.
00:00:59
John S
Yes.
00:01:02
John S
Oh my God. Okay. Is white light and from our ingress of vision the same thing?
00:01:07
johngrimsmo
No, they're actually quite different things. they
Understanding White Light Interferometer
00:01:09
johngrimsmo
They have some related components, but one project turned into two projects.
00:01:09
John S
Oh my God.
00:01:13
johngrimsmo
And then, yeah, the past few weeks have been a total green light of like, okay, I'm allowing myself to do this.
00:01:14
John S
Interesting.
00:01:19
johngrimsmo
And I've spent a lot of time on it. It's amazing.
00:01:24
John S
Cool.
00:01:24
johngrimsmo
Yeah. So, the I mean, the quick version, white light interferometer traditionally lets you measure surface finishes insanely close with nanometer level Z precision.
00:01:37
johngrimsmo
um But it can also measure distances, features, holes to holes, size of holes, things like that. But the viewing the field of view is like like a quarter, but like ah an American quarter, like very small, um you know, half inch diameter kind of thing, 12 millimeter field of view.
00:01:46
John S
Okay.
00:01:53
johngrimsmo
Depends, depends on your lenses and how big you want to see and all that stuff. But it all comes down to how many micron per pixel, like how many physical microns that you're looking at of space
Telecentric Lenses and Vision Systems
00:02:05
johngrimsmo
fit in in one pixel of camera view and different lenses compound or contract that ratio.
00:02:05
John S
That's crazy.
00:02:09
John S
Okay.
00:02:13
John S
Okay.
00:02:14
johngrimsmo
So imagine if you zoom out and you have this wild wide wide field of view, you might get, you know, 12 microns per pixel. So you can't see tiny, tiny details.
00:02:22
John S
Yeah, sure, for sure, sure.
00:02:23
johngrimsmo
but if you zoom in real close and you see this tiny little two millimeter square you might get one millimeter per pixel and the light is half a micron like so you can't you can't zoom in from much more than that so it's pretty crazy so i'm super excited about that i ordered all the stuff i need i ordered a lens a telecentric lens i learned all about telecentric lenses they're really fascinating i ordered a camera from thor labs i got my calibration, calibration square from Thor labs, which is like a precision, um, photo chemically etched, uh, piece of glass that is, has exact 10 millimeter square on it.
00:02:34
John S
needs It's hard to... Yeah, no worries.
00:02:52
John S
Oh.
00:03:05
johngrimsmo
And then a five millimeter square and a two millimeter square and a one and ah all the way down to a 10 micron square.
00:03:05
John S
Okay.
00:03:10
John S
Oh, no kidding.
00:03:11
johngrimsmo
So you can see the optical resolution of your system.
00:03:12
John S
That's awesome.
00:03:14
johngrimsmo
And also with this system, I can scan it and measure it.
00:03:15
John S
Sure, sure.
00:03:18
johngrimsmo
And I can like calibrate my results being like, I know that's 10 millimeters. I know that's, you know, 10 micron. Am I getting the same results from my measurement system?
00:03:26
John S
Yes.
00:03:26
johngrimsmo
Yes. So sick. So that calibration, a little one inch piece of glass, flat plate of glass was 165 bucks And sixty five buck us it's like, it's a lot, but
00:03:33
John S
Oh, okay.
00:03:36
John S
That's all right, but it's not.
00:03:38
johngrimsmo
it's the gold standard it's your it's your ceramic gauge block it's your you know it's yeah it's the the thing exactly um so that's super cool and uh that evolved into once i ordered everything and i was like sweet now what
00:03:39
John S
Yeah.
00:03:42
John S
Yeah, sure.
00:03:45
John S
don't Don't drop it.
00:03:59
johngrimsmo
And I guess I just wait till it all comes in. There's a bit of software stuff I can do in the back end, but it's not as exciting until you have parts and it's like ready.
00:04:07
John S
Yeah.
00:04:07
johngrimsmo
So that kind of left, you know, a five second void of like, well, now what? i was like, Well, what if I, because I can only, my field of view is only so small. What if I make an XY table that can move and stitch it all together? Is that possible? Absolutely. hundred percent. No problem.
00:04:23
johngrimsmo
And that led to, I'm sure we've all seen at trade shows, vision systems where you have a white plate, like a white table, backlit table, camera on top. You put your part on top and it sees it perfectly. Keyance makes
Automating Vision Measurement with Python
00:04:35
johngrimsmo
them. It makes them. Steric makes a nice one. Um,
00:04:38
johngrimsmo
I've seen them. I've never been too excited. i know it's used a lot in the watch parts industry and and various things because you need clear OD profiles or holes.
00:04:49
johngrimsmo
It doesn't measure depth.
00:04:49
John S
Yeah, sure.
00:04:50
johngrimsmo
It doesn't measure weird stuff, contours like vertical contours.
00:04:53
John S
Right.
00:04:54
johngrimsmo
So it never super fit, but we're kind of on and off struggling with two of our parts that are these through holes and whether they're they're growing or shrinking in heat treat or whether tools are wearing and features are small.
00:05:02
John S
Okay.
00:05:07
johngrimsmo
and While we have the CMM, it's a pretty high barrier to entry to like write a quick program and like scan every part.
00:05:14
John S
Yeah.
00:05:15
johngrimsmo
It's time consuming. It's all that. So I kind of was like, well, now that I've learned all this about lenses and and white light interferometer and all that, like what am I missing to make a vision system?
00:05:28
johngrimsmo
Like nothing. like I've already learned all about camera systems and all that. So I ordered a second telecentric lens locally from eBay actually. So that was nice for a couple hundred bucks.
00:05:40
johngrimsmo
And that came in, it was a lot bigger than I expected. it's like a huge rifle scope. You know, it's like 13 inches long, maybe four inch diameter on the end.
00:05:44
John S
Ooh!
00:05:48
John S
Okay.
00:05:50
johngrimsmo
But being a have you ever heard of a telecentric lens? Do you know what I'm saying?
00:05:53
John S
I just Googled it.
00:05:53
johngrimsmo
Exactly.
00:05:54
John S
No, I don't even know. Maintains a constant magnification regardless of the object's distance.
00:05:59
johngrimsmo
Yeah.
00:06:00
John S
What?
00:06:01
johngrimsmo
So I've seen videos on YouTube about this and I'll make my own. Imagine if you have two identical sized objects, like two inch square blocks, and you you shift one further away, like five inches away, they will look the same size to the camera lens.
00:06:09
John S
Yeah.
00:06:12
John S
Yeah.
00:06:16
John S
That's voodoo heresy.
00:06:17
johngrimsmo
they They have zero zero depth of field. So as long as they're within the focal range of the camera,
00:06:21
John S
in Insane. Okay.
00:06:24
johngrimsmo
they will be the same size. And, Edmund's optics has this really neat, um, YouTube video about a model train with a regular camera and a telecentric lens camera.
00:06:26
John S
Crazy.
00:06:35
johngrimsmo
And you cannot tell that the train is getting closer or farther away. It just goes back and forth on the train on the track.
00:06:39
John S
Insane. Oh my God.
00:06:43
johngrimsmo
And it's, it's like, I'm trying to figure out the best way to just at Edmund's optics.
00:06:45
John S
What's that called? want to look that Edmonds?
00:06:47
johngrimsmo
Yeah.
00:06:48
John S
Okay. going to look this up.
00:06:48
johngrimsmo
Um, it's like, in the field of view which depends on the size of the lens and all that stuff it's like there's a it takes just takes a forward-looking slice and every pixel just looks straight forward and it doesn't have the parallax error or that our eyeballs have that like spreads out and like sees movement and change and distance and all that stuff just wild anyway what what
00:07:02
John S
Yeah.
00:07:13
John S
Sure. Sure. Binocular. i mean We have binocular vision.
00:07:17
johngrimsmo
yeah and and even a single eyeball has this like parallax kind of crooked eyeball look so to prove that i took one of our little lock bar inserts and i put it on our zeiss microscope and i it's got a you know microscope optics and a camera on top and i kind of zoomed in and the holes off to the left look crooked
00:07:17
John S
<unk> here
00:07:21
John S
Oh, fair enough. Yeah, sure, sure.
00:07:38
John S
Yeah, sure, of course, has to.
00:07:39
johngrimsmo
And the hole on center, it looks on center and that's how lenses work. But a telecentric lens fixes all that. So everything within the field of view is looking straight down, head down.
00:07:48
John S
His head's drowned. What?
00:07:49
johngrimsmo
Yeah, it's wild.
00:07:50
John S
so so So if you had a revolver cylinder, you know, it's gonna, every bore would look, no way.
00:07:54
johngrimsmo
Yeah, yep. You'd see all the way to the bottom, you would see six cylinder, would look like circles, including the outside profile would be inside and outside profiles would be perfect circles all the way down.
00:07:59
John S
as That's trippy, dude.
00:08:03
John S
Yeah. Oh, that's cool.
00:08:05
johngrimsmo
If if it's all square and aligned and stuff.
00:08:07
John S
Yeah.
00:08:08
johngrimsmo
So my lens came in, it's wild. I'm waiting for an adapter that'll come tomorrow to hook my camera up to it. And yeah, so one microscope turned into two microscopes that I'm building, the interferometer and now the vision system.
00:08:23
johngrimsmo
And vision system for measuring like flat parts with holes in them pretty much. Interferometer for measuring very close features, surface finishes, scratches, depth of grinding marks, all this stuff.
00:08:35
John S
Yeah.
Camera Calibration and Accuracy Improvements
00:08:37
johngrimsmo
And I'm like, so excited for this.
00:08:40
John S
That's good.
00:08:42
johngrimsmo
With the vision system, I wrote an entire Python software with my first user interface, my first program GUI, which is beautiful and amazing.
00:08:49
John S
No kidding, yeah.
00:08:52
johngrimsmo
And I put my little logo on it. um And the goal is You bring a part you want to measure, say a lock bar insert, into the field of view. And once it stops moving, the camera automatically captures it, uses the OpenCV library to notice that, oh, that's a lock bar insert based on your previous test.
00:09:11
John S
Okay.
00:09:11
johngrimsmo
And based on a very quick teaching program, I want to measure that hole, that hole, that hole. I want to measure hole to hole of these two features, hole to hole of those two features, and the angle between these two things, and the length of that line.
00:09:22
johngrimsmo
And the teaching process, I had Leif doing this without any instruction.
00:09:26
John S
John, nuts, right? i mean, I give you a ton of credit because there's a difference between talking about this and actually doing it. So yeah, you're using cursor, but still you're doing it.
00:09:31
johngrimsmo
Yep. Yep.
00:09:35
johngrimsmo
Exactly. and And so the program has the camera tied into it, live view, and you snap a quick picture and then you can now zoom in to the pixel perfect.
00:09:42
John S
Yep.
00:09:45
John S
Mm-hmm.
00:09:46
johngrimsmo
and like And then I'm using the Thor Labs calibration square to like calibrate my view. to make sure that my measurements are accurate. And the goal of this is to is to be able to measure a pretty much one tenth difference in hole diameter on some of our parts, which we stick.
00:10:03
John S
nice. So the the camera is
AI Tools in Manufacturing
00:10:06
John S
what here? The Thor Labs camera?
00:10:07
johngrimsmo
so I got the Thorlabs camera for the interferometer and I learned that a black and white monochromatic camera will be more accurate than a color camera because of the way that cutcent color sensor pixels work.
00:10:17
John S
Okay.
00:10:21
johngrimsmo
Monochromatic gets you more pixels because you're not wasting blue and red or whatever.
00:10:26
John S
Yeah, my understanding is that color CCD just has to alternate red, green, blue.
00:10:27
johngrimsmo
Something about that.
00:10:31
John S
So you're you have three pixels per one output of resolution, whereas in black and white, I'm assuming is just two.
00:10:32
johngrimsmo
Yep.
00:10:37
johngrimsmo
Exactly.
00:10:38
John S
so okay, 30% better, which matters, I guess.
00:10:38
johngrimsmo
Yep. Something like that, yeah, yeah even more.
00:10:41
John S
Yeah, okay.
00:10:42
johngrimsmo
um So your micron per pixel increases like significantly going to a black and white camera, so that's great.
00:10:48
John S
Yeah.
00:10:49
johngrimsmo
And then I got some eBay used, I think they're machine vision cameras, but also like security system cameras from eBay for like 50 bucks for the vision system.
00:11:01
John S
Okay. The open still open CV.
00:11:02
johngrimsmo
now Yep, open CV thing, and mounts right to the top of the telecentric lens with the adapter that's coming tomorrow.
00:11:04
John S
Yeah.
00:11:09
johngrimsmo
And then like my software is, it's not done, but it's workable.
00:11:13
John S
Yeah.
00:11:14
johngrimsmo
So once the I had a testing with a webcam, which has like horrible vision comparatively,
00:11:20
John S
Sure, sure.
00:11:22
johngrimsmo
but like it works, everything works. It's super cool.
00:11:25
John S
So number number one, awesome. I love this. Number two, um details.
00:11:28
johngrimsmo
Yeah.
00:11:31
John S
I'm curious. So you have a, the camera isn't just an HDMI output. So you have it feeding into a, oh, even easier.
00:11:36
johngrimsmo
Yeah, USB.
00:11:38
John S
Okay. So it's just going into a Linux computer
00:11:39
johngrimsmo
Yep.
00:11:41
John S
or
00:11:42
johngrimsmo
Windows.
00:11:43
John S
You're running, okay. So a little harder to run, well, sorry, it's not hard.
00:11:44
johngrimsmo
Yep.
00:11:46
John S
Well, I guess it's harder. It's easier to run Python stuff on Linux or even a Mac book. It's like comes in pre-installed, but okay. um
00:11:52
johngrimsmo
Yeah.
00:11:53
John S
Sick, that's awesome. And so that's what I think is so cool about this. It feels like a ah real big life hack is that the camera is not commensurate without, it's like a crappy, low quality, low cost camera, relatively speaking to what you're getting for measurement capability.
00:12:06
johngrimsmo
Yep. Exactly, exactly. And i i learned all about the, do you want a 4K camera or do you want a, you know, sometimes more is not good. Sometimes more is bad. And for the interferometer, I i don't have this huge like megapixel camera.
00:12:25
johngrimsmo
um I think I got a, forgot it was a five megapixel camera when you can get up to 20 because there are benefits to going smaller.
00:12:31
John S
Okay.
00:12:34
johngrimsmo
um forget exactly.
00:12:35
John S
Smaller, 20 is smaller. oh
00:12:37
johngrimsmo
20 is bigger. 20 makes a huge picture, like 5,000 pixels wide.
00:12:40
John S
ah Interesting. Okay.
00:12:42
johngrimsmo
Whereas my, exactly, my pictures are going to be 1,400 pixels wide, which is not a huge photo, but I'll get excellent resolution and exactly what I want and need from that.
00:12:42
John S
Total megapixels.
00:12:47
John S
Okay.
00:12:53
John S
Yeah.
00:12:54
johngrimsmo
And then it'll process faster.
00:12:54
John S
I love this.
00:12:56
johngrimsmo
Yeah, it's great.
00:12:57
John S
Oh, interesting. Sure, sure, sure.
00:12:58
johngrimsmo
Yeah.
00:13:00
John S
Dude, and like...
00:13:00
johngrimsmo
So it's all coming together. And then I programmed most of the vision system stuff in CheckGPT.
00:13:07
John S
Okay.
00:13:07
johngrimsmo
or use it using chat's help. And then I got, and don't know what the exact rule is, but in the code editing side of ChatGPT, there is a limit to the file size.
00:13:19
johngrimsmo
And for me, it's about a thousand lines of code, anything more than that.
00:13:19
John S
Interesting.
00:13:22
johngrimsmo
And it starts to like break and yell at you. um
00:13:24
John S
Yeah.
00:13:26
johngrimsmo
And it's not really lines of code, it's a amount of code, but whatever. For me, it's a thousand lines of code. And after that, once my code got any bigger than that, i had to i could only paste in chunks of code that it could reference.
00:13:36
John S
Uh huh.
00:13:37
johngrimsmo
And that gets really annoying really fast. So then I looked into other solutions and I remember you told me about cursor a couple months ago. So I looked into Cursor, I looked into Claude, some of the other ones, and I just went for Cursor.
00:13:47
johngrimsmo
And within within half an hour, i bought the full subscription, like $20 a month or whatever.
00:13:47
John S
Yeah.
00:13:53
johngrimsmo
And I'm like, this is the greatest thing ever because it sees the entire code and it just makes all the changes I need.
00:13:53
John S
Yeah.
00:13:58
johngrimsmo
And it's conversational programming. And it's like the coolest thing ever. And it's I'm finding it a lot better than ChatGPT, faster and more responsive for editing programs like this.
00:14:05
John S
Good.
00:14:09
johngrimsmo
And then a couple days later, I tied it into the way we do our buyer's choice system. We have a big git GitHub repository, which I don't really understand how it works.
00:14:20
johngrimsmo
but But Cursor can tie in directly to the entire website of everything we have on GitHub, see all the files, all the everything, and can make any changes that I want conversationally.
00:14:26
John S
ah hu
00:14:32
johngrimsmo
So I made some pretty significant changes to the photo gallery in our buyer's choice order fields. using cursor and GitHub and I push them to GitHub and I'm like, I don't know how this works.
00:14:39
John S
Yeah.
00:14:42
johngrimsmo
So i'm asking like cursor, like how do I actually push and merge and all this stuff? And it just works. And now it's like massive improvement.
00:14:51
John S
Incredible, John.
00:14:52
johngrimsmo
time it's It's been a fun few weeks of total nerdville.
00:14:57
John S
The, I know we've got a very much recurring theme about DIY versus, you know, pay to play or buying stuff. But like, I'll tell you there's just been no innovation in the manufacturing and metrology world.
00:15:11
John S
Well, I always hesitate to make blanket statements, but basically I love what you're doing. I love that it's like, wait, no, we can do this. You don't necessarily need to be spending mid to high five figures from a legacy German or Japanese company to do stuff now that um At the risk of riding on your coattails, it's kind of what I'm planning to do, but have no street cred as of today with with home assistant, frigate, the cameras to like monitor the automation system, trigger things on and off, look at and evaluate the situation.
00:15:31
johngrimsmo
Yeah.
00:15:34
johngrimsmo
Mm-hmm.
00:15:39
John S
um What you're doing is even cooler though, John.
00:15:40
johngrimsmo
All possible. Yeah.
00:15:41
John S
Like it's awesome.
00:15:42
johngrimsmo
Yeah. Thank you. And it's like, I'm...
00:15:44
John S
Yeah.
00:15:46
johngrimsmo
reinventing the wheel, like these things exist, but like you said, they're 50 to $150,000. And im like, no, not right now. And i was thinking about it.
00:15:54
John S
Yeah. Yeah.
00:15:55
johngrimsmo
I was like, if I develop these and they work for our company and everything works, and if I disappeared and the company still needed that technology and mine broke or whatever, my company would now know to just go buy the right one.
00:16:08
johngrimsmo
Like, like we need that. Do you know what I mean by that?
00:16:10
John S
yeah yeah Yeah, but no, like, that's the thing that's... um this is
00:16:14
johngrimsmo
but if If it breaks, that's my question. if i'm If it breaks and I'm not here to fix it, what?
00:16:18
John S
Just use ChatGPT. Yeah. Totally fair,
00:16:20
johngrimsmo
Yeah. Not everybody is like capable or smart enough to do that. I know it seems easy now to you and me, but I'm really starting to realize it's easy to nerds who want to make it happen.
00:16:27
John S
yeah.
00:16:29
johngrimsmo
My brother's never going to do this. You know?
00:16:31
John S
totally totally fair yeah That's a fair point, but it's, I think that will continue to happen and shift and, um, see the light and it gets easy. i mean, we are still in the first inning of like back in the day when you, to have a computer, you had to buy an 80, 80, you know, from a magazine and start assembling circuit boards and, um, yeah.
00:16:43
johngrimsmo
Totally.
00:16:51
John S
Yeah.
00:16:51
johngrimsmo
But yeah, this past few weeks has completely opened my eyes, not just to cursor, but to the potential of like actually using Raspberry Pis and Arduinos and not being so stuck on the code side of it.
00:17:04
johngrimsmo
Because I have zero interest in learning the intricacies of Python.
00:17:05
John S
Yeah.
00:17:08
johngrimsmo
I want to solve a problem. And I respect programmers like nothing else, but I don't care. I just want my solution. And so what I find, like we've had, we've hired many programmers over the years from Upwork and other places, and they do great work, but it's challenging and time consuming to work with them to do what I want.
00:17:17
John S
Yeah.
00:17:26
John S
Yeah.
00:17:27
johngrimsmo
And having cursor or chat or anything, it's like having a pseudo expert sitting right next to you and you're just constantly bouncing back and forth like, okay, i want to do this.
00:17:33
John S
Hundreds.
00:17:36
johngrimsmo
Is that possible? No, you can't do that. But you can do it like this. Oh, sweet. I want to do that. Okay. Next, next, next, next, next.
00:17:40
John S
Yep,
00:17:41
johngrimsmo
Within minutes, you're iterating five times. And you can test it live. That's the beauty with programming. And I'm like, okay, it works great. Thank you. Bye. Send.
00:17:50
John S
yep. I was thinking about that this morning about like, boy, the changes that will come of things, right or wrong. You know, it's kind of like you and I were laughing about your wife, Megan, her disdain for this, but it's like, I'm not here to, I'm just observing what's happening.
00:18:01
johngrimsmo
Yep. Totally.
00:18:03
John S
Like if you're a translator, tough, tough business these days, like a language translator, but also like for the stuff that you and I are doing with the programming stuff, I don't think,
00:18:14
John S
Um, you and I will run into the problems that we're, you're hearing about from like consumer facing software companies, or even like major consulting firms or banks where it's like, uh, there's some hazard to like not knowing what you're doing and building really deep, extensive codes, programs.
00:18:30
John S
Like you and I are building many apps to do like our one task.
00:18:34
johngrimsmo
Yeah.
00:18:34
John S
Like it's probably, I think we'll be okay. Um, yeah.
00:18:37
johngrimsmo
Yeah. So I kind of hesitate, you know, telling the world preaching from the rooftop that like AI cogeneration, greatest thing that's ever happened, but it kind of is.
Integrating Metrology and Automation
00:18:48
John S
Pretty awesome, John. um
00:18:49
johngrimsmo
Yeah, it's pretty awesome.
00:18:51
John S
Well, the you talk about that glass square reminds me of top of mind here of wanting to integrate integrate metrology into this new automation cell. And I will always remember the anecdote about the Renishaw Equator, which is not a CMM because it doesn't really have absolute accuracy as relative
Vision Systems in Quality Control
00:19:13
John S
accuracy.
00:19:13
johngrimsmo
Okay.
00:19:13
John S
So I believe you have to repeatedly or not repeatedly, but you have to often recalibrate it on a feature, like a master that's so sort of like, okay, if you're telling me that this ring is one inch, I can measure that.
00:19:24
John S
And then I can go measure your part.
00:19:25
johngrimsmo
It's a comparative.
00:19:25
John S
That's one inch and tell you, yeah, which is kind of fine for a lot of our production stuff.
00:19:26
johngrimsmo
Yeah. Mm-hmm.
00:19:31
John S
We're not a job shop. We don't have part variety, so to say. But even a Renishal Equator, which ah years ago was probably 30 grand, I'm kind of like, eh, what do we look for for, you know, a system like this where you can do something? And sometimes we're trying to do, like, here's a vision, like whether it's OpenCV or Frigate, like, hey, take the, out of every batch of 50,
00:19:56
John S
let's ah so I'm assuming a batch size of 50 for no reason, but if there's 50 mod vices, one goes through a full personal QC, but one goes to the vision system that just looks at the side to see, hey, is the end mill, you know, and we'll wear and you'll have like a notch and it doesn't affect it. Most customers don't even notice it, but you know,
00:20:15
John S
how so its you know It depends on my mood or the customer's mood. Like, do you care about the or the machine? Like, what do you want to do? Do you want to change the tool? Like now all of a sudden you can start documenting that, saving logs of it.
00:20:26
John S
The holy grail, which I don't think this will happen in our lifetime, would be you take something like chatter where it could own all information on the machine and it knows, okay, tool 41 was replaced, you know, 70 or it's been in the cut for 13 hours.
00:20:39
John S
Now this is what the service finish looks like and make
Using AI for Complex Programming Tasks
00:20:42
John S
its own decisions.
00:20:42
johngrimsmo
Yeah. Oh,
00:20:43
John S
Yeah.
00:20:46
John S
The only other thing I'll mention is, um and we might've talked about it, but the friend who's building the eyeballs with Python and all the different stuff.
00:20:54
johngrimsmo
Yep.
00:20:55
John S
Oh, sorry. That's triggered a conversation with Alex here who was saying that chat GBT has a like a version within it called codex.
00:21:05
johngrimsmo
Okay.
00:21:05
John S
do Do we talk about this?
00:21:07
johngrimsmo
No.
00:21:08
John S
So what I had sort of said to Alex was, hey, the thing that I don't know about using an LLM to program is that like, okay, it can make like a single Python file, like the prime number generator. It's, you know, 50 lines of code, whatever.
00:21:21
John S
The ah Mike who's building the eyeball thing has, you know, seven different Python files that do different things. And like, that's where I quickly betray my ignorance as a programmer I don't understand when and why you're having different files and, you know, what do they do and how do they talk to each other?
00:21:34
johngrimsmo
Right.
00:21:37
John S
and I was sort of saying this to Alex and now i was just like, oh no, you've got to try codecs within ChatGPT. It can do all of that.
00:21:44
johngrimsmo
Nice. i know I know Cursor can as well. So maybe the maybe the codex is Chat's version of that.
00:21:47
John S
Okay, there you go. Yeah. Yeah, sure.
00:21:51
johngrimsmo
Interesting. I hadn't heard about that. Chat has the Canvas, which is your code editing side. Um, but that's what I'm finding the limit of like a thousand lines of code or whatever until it just goes, I, that's the limit.
00:21:58
John S
Okay.
00:22:04
johngrimsmo
I can't see it all anymore.
00:22:06
John S
Yeah. Where, where I, um, I did a bunch of traveling, which I'm happy to talk about. And one of the things when I was super inspired and fired up was I sat in an airport for two and a half hours and I tried using both chat as well as, is it Anthropic?
00:22:21
John S
Is that the, like blanking on the name?
00:22:22
johngrimsmo
I've heard of it, but I don't know what it is.
00:22:25
John S
Um,
00:22:28
John S
is it called Claude? Yeah. It's the Claude software, which is, which is also held in high regard, um, to create fusion, um models.
00:22:29
johngrimsmo
Okay.
00:22:39
John S
And the, the chapter is not closing this. i actually have a zoom with somebody that I want to learn more because they're doing it. And I'm, I was sucking wind big time. So, What I said to Claude was, I want to create a set of soft jaws for a curt six inch vise that will hold two one inch parts, a quarter inch deep.
00:23:02
John S
And I thought, okay, that's a reasonably complicated thing
00:23:04
johngrimsmo
Yeah. If you told a machinist that, they'd have enough information mostly to figure that out.
00:23:10
John S
Bingo. But like typical, my like Luddite septicism, like thought that they wouldn't, wouldn't know how to do it. Dude, nailed it. It was like current six inch vise uses this size jaws. I'll grab these standard soft jaw size. We'll do this pocket here. I'm assuming you want it centered. Like it just nailed it.
00:23:27
John S
But then where it failed and it failed, the whole time was the Python file that it output, you then go into Fusion and you run that as a script, it would never build the B rep.
00:23:36
johngrimsmo
Okay.
00:23:39
John S
um and And I say this earnestly, I spent two hours all the way going back down to a simple cube
00:23:46
johngrimsmo
Yeah. Yeah.
00:23:47
John S
And it still kept failing. And I was like having this dialogue with it. And I'm like telling Claude, like, Hey, you know, it's funny how you, uh, treat it like a human. You're like Hey, I'm not trying to be mean, but you keep saying you can do this and that keeps failing and you're wearing out my patience.
00:23:58
johngrimsmo
Yeah.
00:23:59
John S
I'm like, can you just create a cube that will not fail?
00:24:01
johngrimsmo
Yes.
00:24:01
John S
And it's like, yep, here's the code. This will absolutely not fail. And I load up fails.
00:24:04
johngrimsmo
yes
00:24:04
John S
Um, but I want to learn more and I did the same in chat. And so I'll come back to everybody when we do learn this more because
00:24:11
johngrimsmo
Okay.
00:24:14
John S
I just, the I'm very curious about this.
00:24:17
johngrimsmo
Absolutely. I mean, in that scenario, it sounds like, like you're missing one thing that's causing it
00:24:23
John S
It was an extrude that was throwing it off. Yeah.
00:24:26
johngrimsmo
Yeah.
00:24:27
John S
And you can start to parse through it. Again, but like, I'm like you, I don't want to learn the Python code to program, but it was not grabbing the correct edge or face to do the extrude.
00:24:30
johngrimsmo
Yeah.
00:24:35
johngrimsmo
Okay.
00:24:36
John S
So you could sort of see how it failed. Anyway, really cool.
00:24:38
johngrimsmo
Yeah. Cool. Really cool.
Insights from Hacksmith Visit
00:24:44
johngrimsmo
um Okay, I got one more thing and then I'll let you go nuts.
00:24:48
John S
Yeah.
00:24:50
johngrimsmo
So yesterday I went to Hacksmith again for the second time and did a full, well, not so much a tour. So they launched, like when Leif and I went there a year ago, they were like, hey, we got this idea to make a knife.
00:25:01
johngrimsmo
Can you give us some pointers or whatever? Then this summer they launched the Kickstarter. They got $14 million dollars in funding 38,000 backers or something.
00:25:08
John S
crazy
00:25:12
johngrimsmo
and um And now they're in production. They bought 32 CNC machines. They said they bought ah a whole bunch of Haas machines and a couple of styles. And they said it was Haas's biggest single Canadian order of CNC machines.
00:25:27
John S
Hilarious.
00:25:28
johngrimsmo
And so now they have one hanger full of, I think it's 12 DTE ones or twos, and then a bunch of other machines sprinkled out everywhere else.
00:25:40
johngrimsmo
And, uh, so was interesting because they're having some QC issues and some assembly issues. And they, they asked me politely they're like, can you come and like, like hold our hand for a little bit?
00:25:48
John S
yeah sure sure
00:25:49
johngrimsmo
Um, and I'm like, yeah, I'm happy to do that. Can I, if I, if I can film, cause I think it'll be really interesting conversations. And they're like yeah, of course. Um, so my videographer, Ryan and I went there yesterday and we spent the whole day there filmed two hours of footage.
00:26:03
johngrimsmo
Um, where we QC knives together and we talked about their finishing processes and their machining processes and tolerances and things like that. And I had a total blast. And I've got a few little videos. i was like, James, I'm going to film for 60 seconds. Say something cool. Like, show me something cool.
00:26:23
johngrimsmo
And he showed me some cool home assistant automation stuff that he built.
00:26:26
John S
Really?
00:26:27
johngrimsmo
so Yeah, yeah.
00:26:28
John S
Sweet.
00:26:28
johngrimsmo
You're like,
00:26:29
John S
Oh, good.
00:26:30
johngrimsmo
So yeah, little things like that. It was is cool.
00:26:34
John S
the I think I watched either the Kickstarter video or maybe just one of their main channel YouTube videos. Those machines were crammed in there.
00:26:43
johngrimsmo
Yeah. Yeah, they're like kissing, like side to side and in front to back.
00:26:47
John S
Yes. Yes. Like, yeah, no aisle, like no hallway, aisle, whatever.
00:26:49
johngrimsmo
Yep.
00:26:52
johngrimsmo
Yeah, basically they're trying 40,000 knives tomorrow.
00:26:52
John S
Interesting.
00:26:56
John S
Yeah. Yeah.
00:26:56
johngrimsmo
and uh on paper that sounds linear but then for one manufacturing is hard for two making knives is hard um assembly is hard and then they told me yesterday they're like you know we we thought the machines and the machining and all the infrastructure and stuff would be the hardest part and we so we put all our effort into that and now it's assembly that's the bottleneck of everything and i'm like yeah because it's a whole picture it's like it it all happens and
00:27:02
John S
Yeah.
00:27:21
johngrimsmo
you know, say your machine parts are too tolerance, that your theoretical engineer says they should be this.
00:27:26
John S
Yeah.
00:27:28
johngrimsmo
And then the assembly guy goes, I can't fit them.
00:27:28
John S
yeah
00:27:30
johngrimsmo
They don't go together or I'm scratching them or there are machining marks that won't tumble out or whatever. All these things that requires the full feedback loop that we've worked so hard to learn for.
00:27:40
John S
Yep.
00:27:41
johngrimsmo
Which is why while we have two buildings and two separate operations, we talk daily about like, oh, still struggling with this. Or if you could fix that, I could fix this. Or my life would be better if if I did that for you. All that stuff. And and they're they're just hitting that wall right now.
00:27:57
johngrimsmo
They're like, we made a whole ton of parts and we can't get any to put together.
00:27:57
John S
Yeah.
00:28:00
johngrimsmo
if We're like, okay, yeah, I think I can help you out a little bit.
00:28:01
John S
Yeah.
00:28:05
johngrimsmo
So... So yeah, um it was cool.
00:28:08
John S
It's cool. That's awesome.
00:28:10
johngrimsmo
It was cool. And we talked briefly before we hit record about maybe adding a little segment to the show called, what do we call it? Business ideas?
00:28:22
John S
We were, yeah, I don't remember it was, it doesn't even need one, but we were laughing and I don't remember whether whether it was on.
00:28:23
johngrimsmo
We don't have a name for yet.
00:28:30
John S
air or right after we hit the, uh, end record and recording button. But you know, it is the business of machining and we have no obligation to talk business. Um, and i in fact, frankly, saying it like that almost makes it sound snoot snotty, but like there are a lots of aspects of the business that we probably just either don't share because sometimes, you know, just like to nerd out on machine stuff, but, um, I thought, well, let's see if we can't find, you know, kind of uh, yeah, just, it doesn't.
00:28:54
johngrimsmo
Yeah, carve out a little... ah little helpful things things that are on our mind things we're thinking about and something hit me yesterday at Hacksmith shop that I kind of went home and sat on and thought about for a little bit they are building a process to make this knife and their whole team of engineers that like engineered it together and love to solve problems with engineering great more power to them
00:28:59
John S
Yeah.
00:29:11
John S
Okay.
00:29:17
John S
Yes.
00:29:19
johngrimsmo
Um, and because they don't already have established processes, somebody has to step up and say, no, we're doing it this way. Like all these new hires are doing it this way. And, and that's fine.
00:29:28
John S
Yeah.
00:29:30
johngrimsmo
Um, But the depth that these nerdy engineers, especially James, want to like go to to solve these problems using Home Assistant or using all these trackers and GPS trackers.
00:29:43
johngrimsmo
And you know even he was talking about having little ultrasonic sensors on every wrist that knows to one centimeter precision like where your hand went to reach into which bin.
00:29:52
John S
Are you serious?
00:29:53
johngrimsmo
Yeah. And like technically, that's freaking amazing, but it's also like a little bit too much for
00:29:54
John S
Oh, holy.
00:29:58
John S
let's build Let's build a knife first and then we can, you can't improve a process that you haven't even, whatever, I'm backseat driving.
00:29:59
johngrimsmo
It was well the night first. Exactly. Right. Exactly. So that got me to crunch on. My business theory for the week is, and I've fallen into this trap before where I know there's a problem in the company and I solve it on my own time at home. And I present a solution to a problem that maybe the rest of the company doesn't see or know about or like feel yet or something like that. But I'm like, Hey, I got the solution. i built this QR code scanning rig so that if we can scan each operation and follow apart throughout the process, sounds awesome.
00:30:35
johngrimsmo
super hard to adopt, like nobody else cared. And it's just adding more work to their lives, even though in my mind, logically, I'm like, this will be sick.
00:30:38
John S
Yeah.
00:30:44
johngrimsmo
This will be amazing. Management can have top level view of what's going on. And then super easy for the operator, you just pick up the gun, boot boop, boop. You know, I tried really hard to make it super easy, but it was pain to adopt, which was demoralizing for me and like for them too.
00:30:55
John S
Yeah. Yeah. It's
00:31:01
johngrimsmo
So
00:31:04
johngrimsmo
watching don't want to say this in a bad way because they don't have processes at hacksmith yet they're building them and it's they're <unk>re you're using these engineering topics to from the start i can't do that so i realized what i have to do is i have to flip it around and i have to be like especially with cursor now i have to tell my team my my leaders to be like guys i can solve these problems now but you have to tell me how you want them solved or or what you want solved and i will so i will find the how
00:31:11
John S
okay. Yeah.
00:31:27
John S
Yeah.
00:31:33
johngrimsmo
If like, I think QR code scanning the order slips, because now we do have paper travelers for every order going through the shop. And I know Spencer and Angelo and Eric are spending probably cumulative one to three hours a day on spreadsheets, just dicking around and like, like moving numbers between all three of them.
00:31:40
John S
Mm-hmm.
00:31:47
John S
Oh, wow. Yeah, yeah.
00:31:52
johngrimsmo
And like, I can solve that. But i want I want them to tell me what they really need and i can solve their problem instead of me solving what I think they need.
00:31:55
John S
Yeah.
00:32:04
johngrimsmo
And then they don't have any buy-in. They don't have any, you know, they can talk to their teams.
00:32:06
John S
Exactly.
00:32:07
johngrimsmo
Everybody can like really feel like I'm solving their problems, not just giving them mystery solutions.
00:32:07
John S
Yes.
00:32:11
John S
Yeah.
00:32:13
johngrimsmo
um
00:32:14
John S
Yes. The top down fallacy.
00:32:16
johngrimsmo
Yeah.
00:32:17
John S
It doesn't, sometimes it's great. Sometimes like, no, this person, manager, boss, teammate, what doesn't matter, has this a passionate idea and it just works.
00:32:25
johngrimsmo
makes the call and does it.
00:32:25
John S
um
00:32:26
johngrimsmo
yeah
00:32:26
John S
Yeah. but Versus sometimes it's just like, dude, we didn't want it. and Now they're forcing us to hit a button and we don't want to hit like whatever.
00:32:31
johngrimsmo
Yeah, exactly. So i yeah, that's what I want to do that.
00:32:33
John S
Yeah. Good for you. It's yeah.
00:32:36
johngrimsmo
I haven't told anybody yet, but that's what I want to do Yeah.
00:32:39
John S
Yeah, that good. I'm not great well, I'm not great at that. I wish I was better. i Sometimes I do it. a lot of times I don't. I'm just like figure out a way to let it be somebody else's idea.
00:32:50
John S
Like steer them toward it's their idea.
00:32:50
johngrimsmo
Yeah.
00:32:52
John S
Give them bind, give them the the tool. Like I would almost say like, you really shouldn't even be the one running the cursor code. Get somebody else to do it. And then you, you know, you're telling them to use cursor or you're telling them, um,
00:33:02
johngrimsmo
more
00:33:05
John S
I just randomly, not tangentially, I just was hanging out with Carl Bass again, the former CEO of Autodesk. And there's a story that's, this is the Wikipedia version of it. So I have no idea how true it is, but it's Wikipedia where he was, i believe let go from Autodesk.
00:33:21
John S
This is in the nineties, perhaps long time ago.
00:33:23
johngrimsmo
Okay.
00:33:23
John S
And the software team um kind of revolted and told the, I guess he wasn't the CEO then, maybe he was like the chief.
00:33:23
johngrimsmo
Yeah.
00:33:31
johngrimsmo
Sure.
00:33:31
John S
CTO, I don't know, it doesn't matter.
00:33:33
John S
They told the CEO, like, we we need his software leadership. be Like, you need to bring him back. And he did, and then he came back. And it's like, yeah they need your team needs a John Grimm's vote to say, here's when you use cursor, here's when you call Zeiss, here's when you, like, you know what I mean?
00:33:47
John S
Like, that's the steering the ship side of it.
00:33:49
johngrimsmo
yeah Yeah. Here's how to solve your own problems kind of thing.
00:33:53
John S
Yeah.
00:33:53
johngrimsmo
Yeah. Yeah.
00:33:56
johngrimsmo
Hmm.
00:33:57
John S
Yeah.
Experiences from Recent Work Trip
00:34:00
John S
Well, cool.
00:34:00
johngrimsmo
So those are my notes.
00:34:01
John S
um
00:34:02
johngrimsmo
Like I've, I've had a good few weeks.
00:34:02
John S
Good.
00:34:04
johngrimsmo
I'm excited.
00:34:06
John S
So I was off um on a work trip to, that started because of a tool path board meeting that was being hosted Autodesk headquarters, which I had not been to in a while. And I've said this before and i I'll continue to say it for ah a you know kid that lives in rural Ohio, being in California in that tech space is just, there's ah there is a contagious energy. I know that's cliche, but it really is so I did a good job at making the most of this trip.
00:34:37
John S
I flew out to Idaho first and met the folks, well, I've met them before, but met the Lights Out Machining, Lights Out Manufacturing folks at their facility.
00:34:40
johngrimsmo
And
00:34:47
John S
they were They had our automation cell almost ready to ah ship ship to us, which it's now in transit. So spent an afternoon with them looking at their UMC 500, looking at their old robo-drill, or older, RoboDrill, looking at some training and learning stuff.
00:35:04
John S
um to To come back to your point, I kind of did an awkward job of being like, I'm happy to learn, but I also, I've realized I get excited and passionate about things.
00:35:16
John S
And so if I learned that you are, then it will smother other people out because I get so excited. And I think it's, for me, it's better to actually keep a distance, which is, which i mean, I'm not looking for sympathy here, but it kind of hurts because it kind of sucks.
00:35:27
johngrimsmo
yeah yeah
00:35:31
John S
But like, um you know, we have a young intern who's doing great. And if I all of a sudden, learn faster and know more than her because guess what?
00:35:40
johngrimsmo
yeah
00:35:41
John S
This is my full-time job. I've been doing this for 20 years, blah, blah, Then she's just like, what do you want me to do Like, it's like, nope, Courtney, you you run with this. Ask me when you have problems. I can solve it with you. i get you in touch with the right person.
00:35:52
John S
I like where that's going. So it's kind of weird to spend time with them, but also be like, don't, I don't want you to teach me stuff, guys.
00:35:57
johngrimsmo
Yeah.
00:35:57
John S
Sorry, not my, um but still really good group. I'm cautiously optimistic about what that looks like. um i'm going to keep moving here because i did a lot. Then hopped over to MetalQuest.
00:36:10
John S
They have an awesome Nebraska shop that I toured years ago with a whole bunch of triple turret lathes and horizontals and B-axis, Akumas and robots and all that. High volume shop, like truly high volume shop.
00:36:23
John S
They opened an Idaho facility five years ago now. It feels like yesterday and in my sort of talking with with Scott, their CEO.
00:36:27
johngrimsmo
yeah
00:36:31
John S
Great guy, great team, great person, great family. And they now have three index, multi-spindle, whatever you call those monstrosities.
00:36:39
johngrimsmo
whoa yeah the machines like that can make a complex part in like eight seconds yeah
00:36:42
John S
um They make a lot of parts that are you know round and conical shaped in the for the firearms world and hydraulics world as well. And they were shown it.
00:36:55
John S
Bingo. Yeah. Yeah. So you obviously typical part example would be a K baffle of a silencer where you're talking about fair amount of OD and ID work and turning work and milling. And they were doing face, they were doing face knurling and ID burnishing with cog steel tools.
00:37:09
John S
um
00:37:09
johngrimsmo
Yeah.
00:37:10
John S
The part that came off was, was actually for the hydraulics world. And it was, um you know, about the size of your, the diameter your thumb about half the length with lots of tight tolerance features like i mean i'm just not saying that um stainless steel like very very respectable part and uh you know if i looked at that part on a typical turning center you're looking at two three four minutes and they're going to know at 11 seconds um it's just right i mean it's just and they're insane um So it's really fun to see their shop. And also, you know, business machine to talk about the business side of it. So he, um Scott splits his time between Nebraska and Idaho. That is not an easy thing to do. Running two shops, two facilities, two teams is great.
00:37:56
John S
devastatingly brutal on a person, I think. He said they made a really good decision where they they treat the the facility like one facility. So there is no, oh, like that's Idaho software or that's like everything's VPN together, everything's tied together.
00:38:06
johngrimsmo
Hmm. Hmm.
00:38:13
John S
And he's really glad they did it that way. I think that's a decision you don't necessarily know until you get downstream. But I can tell you from even running our training facility for a while across a parking lot, I hated it. It's like, oh, I don't have that pin gauge set over here. I don't have those mics over here. I don't have that material here. So how they do it, I give them a ton of credit.
00:38:31
John S
um Flew to San Francisco, got to see Dave Precise. He's got a new startup. They're working on AI stuff and automation and that kind of thing. super, super trippy. He, he convinced Willemann to sell him. Well, don't know if convinced him. He, he, he has the sister to our old Willemann.
00:38:50
johngrimsmo
No.
00:38:50
John S
The machine, the machine that went to Marcus, the 408, that was the one serial number off from 2003 or four was sitting in his shop. I'd never seen it with
Shop Updates and New Machines
00:39:01
John S
my own eyes before. I'd seen it on Instagram.
00:39:04
John S
And they had it. So it was super funny to see that. And they just got into VF2, so they're tooling that up. um And we went to dinner with his, a couple of his guys, software guys, machinist guys.
00:39:14
John S
So you're just like randomly in in Berkeley, you like all talk and shop. And then the Input Inc. kind of main ah software guy, or excuse me, SolidWorks guy um for the Johnny Five build.
00:39:28
John S
I got a chance to meet him for the first time. his he lives...
00:39:31
johngrimsmo
It's...
00:39:32
John S
stone's throw from where we were having dinner. He came over that whole crowd intermingled and we talked about everything from manufacturing to Johnny five. So that was awesome. Then we had the Autodesk board meeting, which was, or excuse me, toolpath board meeting at Autodesk. That was great. Got to see peer nine again and the machines there. Talked to other machinists. Like that was great. um Came home, was here for a couple of days and then went to,
00:39:53
John S
a family member moved to the Great Plains region. So we flew to Dallas for half vacation, half work, where got to meet Tyson Lamb at his shop. I'd never actually seen his shop.
00:40:07
John S
So in Tyson, you he reminds me a lot of you, John, like his artistic vision and the way he creates a a consumer facing product is just ah yeah unbelievably respectable.
00:40:12
johngrimsmo
Yeah.
00:40:17
johngrimsmo
Him and I were emailing before Instagram was a thing.
00:40:20
John S
Is that right?
00:40:21
johngrimsmo
Yeah.
00:40:21
John S
That's funny. Yeah. Yeah.
00:40:22
johngrimsmo
Back when he was still in the garage and I was still in the garage. We have these old emails.
00:40:25
John S
Yep. Yep.
00:40:26
johngrimsmo
Yeah.
00:40:27
John S
Yep. And I think he's got tons of room to grow in terms of how he runs the shop side of things. um
00:40:32
johngrimsmo
Yeah.
00:40:33
John S
But loved, it's awesome to see him. He was so kind with the kids. Like William Jean got to stamp their own golf ball marker um and just had a good time doing that.
00:40:39
johngrimsmo
Nice.
00:40:44
John S
And then got to meet the other a gentleman in Texas who has a truly finished J5. Like he's going to Comic-Cons. I've gotten to know him again online, but first time I met him and hung out in his place for three or four hours, seeing his robot, talking about everything. It's like super weird to see one.
00:41:03
John S
Another one?
00:41:04
johngrimsmo
Yeah, like done.
00:41:04
John S
Because, like, I've spent... Well, yeah, thanks. Jane, ah ah oh, my gosh, she was so excited. She's like, his name's Sean. He's like, Sean, how long did it take you? So Sean's like, ah four or five years.
00:41:14
John S
She's like, well, my dad has spent eight years on it, and he's still not finished. I'm like, thanks, thanks, Jane.
00:41:19
johngrimsmo
It's sass.
00:41:20
John S
Yeah, have this ass. Um, no, so that was, it was, uh, uh, a good, but whirlwind bit of travel. And then now, um, back here, Saunders and we have, um, all the machines except for the horizontal, which I'll talk about in a second, all the machine stuff's done. So the lathe is gone. Saunders is lathe free.
00:41:41
johngrimsmo
nice
00:41:42
John S
Asterix hashtag, we have a VVOR bench not believe in somewhere in the corner, but that doesn't count. um Late is gone, UMC 350 is gone, and we got the Brother S700 XD2, it's here.
00:41:53
John S
And then we got the UMC 400 yesterday, ah the same day we had five inches of snow in the ground, but it was went fine.
00:41:56
johngrimsmo
Nice.
00:41:59
johngrimsmo
Yeah.
00:42:01
John S
Both the the brothers installed, but they do training next week. The Haas is here hooked up, but they haven't yet come. It got came a yesterday. they haven't come and done the install yet. So all those pieces are coming together, which is the plan.
00:42:12
John S
Like I wish between you and me, I wish we were further along in the sense that like I'm anxious, but like no, this was the plan. Get through all this stuff um and robots and root.
00:42:18
johngrimsmo
Mm-hmm.
00:42:23
John S
Yeah. Come together.
00:42:25
johngrimsmo
Amazing. So I saw the UMC 400 come in and I was like, i don't know if I knew about that one.
00:42:31
John S
Oh, really?
00:42:33
johngrimsmo
I knew you had the 350 and I didn't know you were selling. i don't know. I forget, lose track, but.
00:42:37
John S
so Totally fair. And actually Dylan asked me on WhatsApp, he's like, why did you buy the 400 over 500? Which I figured that's a good time to address it. So we, um the 350 being totally honest does have a like stepchild reputation for the reason that Taas released it and then discontinued it.
00:42:57
John S
I can only tell you that we had no problem with it. It ran in fine. We made made perfectly good parts with it. Yes, it's a lighter duty five axis that's a single support thing.
00:43:08
John S
Like it's not a Hermla, whatever.
00:43:09
johngrimsmo
Yeah.
00:43:10
John S
But ah I did not have that criticism. I would have happily kept that machine as a, you know, whatever. But I did want 30 tool count. I did like the idea of a better a machine that they've redesigned new spindle, better rigidity, blah, blah, blah.
00:43:22
johngrimsmo
Mm.
00:43:24
John S
And, um, the, uh, integrated platter versus a T slot that we lose an inch. And we got the auto door on it. None of these were necessary, but basically I was like, Hey, if I can sell the UMC 350 for this, and I can buy a new one for this, the Delta is this.
00:43:39
John S
And I like that I'm willing to do that. And I, we pulled it off.
00:43:40
johngrimsmo
Yeah.
00:43:42
John S
Um,
00:43:42
johngrimsmo
Nice.
00:43:43
John S
ah The why we didn't get a 500, that is a lengthy answer, um which is gonna be a specific question from anybody who's buying a machine, but the 500 is a better machine full stop.
00:43:58
John S
It is more rigid, it's supported on the front of the it's a dual support trunnion, if you will, much higher table capacity. The 400 is still pretty light duty in terms of what it can, how much weight you can have on the table.
00:44:10
John S
um
00:44:10
johngrimsmo
Yeah.
00:44:11
John S
but we had a 500 and it's a big machine. The 400 is effectively a D.
00:44:15
johngrimsmo
you You got a 750 back in the day.
00:44:17
John S
Yeah, we did. Well, the 500 doesn't feel that much smaller than a 750.
00:44:20
johngrimsmo
Yeah.
00:44:21
John S
So everyone needs to do their own research. um But for us, I wanted a small machine, R&D, you know, small parts.
00:44:30
johngrimsmo
it
00:44:31
John S
Like I'm, i'm and well, I haven't run the machine yet, but I still think I'm happy with this decision. i just know i didn't want a big 500 anymore.
00:44:37
johngrimsmo
Yeah, sure.
00:44:39
John S
um
00:44:40
johngrimsmo
Yeah.
00:44:41
John S
And we looked at a brother u five hundred And i'd I'd like to like that machine.
00:44:44
johngrimsmo
Yeah, I was wondering about that.
00:44:47
John S
I'd never got a quote from Yamazin, but I know anecdotally, it probably was going to be about twice the price of the 400.
00:44:54
johngrimsmo
Really? They're about 200,000, something like that, aren't they?
00:44:56
John S
Yeah.
00:44:57
johngrimsmo
Yeah.
00:44:58
John S
Comment stands.
00:44:58
johngrimsmo
Yeah, really.
00:44:59
John S
oh
00:44:59
johngrimsmo
Okay.
00:45:00
John S
So it's just like, look, I'm at that point now in my life where i'm ah i'll take a I'll take a risk on the 400 to see if it does what we want it to And if it does, great.
00:45:08
johngrimsmo
Hmm.
00:45:08
John S
and If it doesn't, I'll sell it and figure something else out.
00:45:10
johngrimsmo
Yeah. It's a 400 really closer to a hundred thousand than, than 200,000.
00:45:17
John S
um
00:45:17
johngrimsmo
Like give or take that's a big difference.
00:45:19
John S
Yeah, John, it is.
00:45:20
johngrimsmo
Yeah.
00:45:20
John S
I mean, look, it's no secret. I won't.
00:45:23
johngrimsmo
Like speedio, our three axis speedio was about, I think one 50 with all the bells and whistles and the door automation and all that stuff.
00:45:23
John S
I'm just like,
00:45:28
John S
Yeah.
00:45:30
johngrimsmo
Um, and I think the, the five axis package is another 50 grand on top of that, give or take.
00:45:37
John S
No, that matches what I heard. um UMC 400, think is somewhere between 125 and 145.
00:45:43
johngrimsmo
Okay.
00:45:44
John S
And I didn't get like, i stopped commenting because we don't, on like partnerships, because we don't we haven't done them and if ever, in years. And so we just, Haas was giving almost anybody who asked a pretty good discount these days.
00:45:52
johngrimsmo
Yeah.
00:45:56
John S
So we just got a nice discount.
00:45:56
johngrimsmo
Nice. Yeah, good.
00:45:58
John S
um
00:46:01
John S
What else I going to say? Oh, okay. Randomly coming back full circle to a fun thing. We found a third UR robot and working out a deal on it used. And you ready for this? You ready where it was coming out of?
00:46:16
johngrimsmo
Toronto, I don't know.
00:46:18
John S
No, no. Cutting tool facility where they had built a test R and D lab mechanism of putting white light interferometers on the end of a UR 10 to measure
00:46:30
johngrimsmo
Shut the front door.
00:46:32
John S
edge prep on coding stuff. And this was a company that already has, i believe dozens of white light interferometer.
00:46:36
johngrimsmo
What?
00:46:40
John S
So they know what they're doing and they were testing out to do it.
00:46:42
johngrimsmo
No way.
00:46:43
John S
And it ends up that the U, yeah, it was really cool.
00:46:46
johngrimsmo
i've I would love more information about this company. Just like see.
00:46:50
John S
It failed. The URs, you need white light interferometer on a like perfectly rigid, infinitely rigid, holding device and they were trying to hold the white lightness property on the end of a UR robot and it just isn't stable enough.
00:47:02
johngrimsmo
Yeah, it just needs to stay still for a few seconds is what the goal is while it scans the part, but it needs to stay like really still for a few seconds.
00:47:07
John S
Okay.
00:47:11
John S
Yeah. And I didn't interrogate the guy like he was just sharing this.
00:47:13
johngrimsmo
Interesting though.
00:47:14
John S
I'm assuming that there's still at a true, ah there's some noise in the service system, harmonic drive, that's
Financial Health and Monitoring
00:47:22
John S
creating oscillations.
00:47:22
johngrimsmo
yeah yep everything like at kern they had the zygo machine which is like the best interferometer out there and there was a speaker next to it playing music and you could see the light bands dancing in the music literally and i will be proving this once i have mine yeah exactly
00:47:22
John S
Even if they're 20 million or nanometers, it's just, it's oscillates.
00:47:35
John S
ha
00:47:38
John S
Exactly. Yeah. Oh, that's Whitney Houston. yeah um Okay. my um The business thing I wanted to share is um There's just the very unprepared, candid, how I look at our historical financials for Saunders.
00:47:58
johngrimsmo
Okay.
00:47:59
John S
um And I don't have a, like, i myself wonder why I look at them. Of course you need to look at them, but like, I don't look at them with, so far with the reason and there or the need to change things. It's not like I say, oh, you know, we're doing so poorly that I need to like find a way to sell more or we're doing so well that I need to like, don't know, you just look at them to make sure everything's okay.
00:48:26
John S
um
00:48:26
johngrimsmo
Now, when you say historical, you mean the year over year or what's the timeframe we're looking at here?
00:48:30
John S
So that's really the punchline of what I wanted to say was it's monthly. So i do the I do the reconciliation every morning myself.
00:48:34
johngrimsmo
Okay.
00:48:39
John S
And I like that because it does help me just make sure mostly we all have any fraud and it takes me so little time.
00:48:45
johngrimsmo
Yeah.
00:48:46
John S
um I could outsource it. It's one of a few things I haven't, but I like just having a little bit of a pulse on that now. It's mindless, et cetera. But then at the end of the month, I export the financials and I do the bank reconciliation to make sure our accounts tie out.
00:48:59
johngrimsmo
Mm-hmm.
00:48:59
John S
At that point in time, I exported an income statement to an offline file where I track our historicals. Sometimes I will make minor adjustments if, for example, and this quickly can get, i don't say it can get complicated. Most people will know, but like, let's say we spent $10,000 on Rego Press.
00:49:18
John S
Uh, technically that should be booked as equipment and it would be subject to depreciation. So it wouldn't show up in an income statement. Um, by the way, if anybody wants to learn more, we did a YouTube video called accounting for machinists, watch it.
00:49:29
johngrimsmo
Yeah. Mm-hmm.
00:49:29
John S
None of this stuff is complicated and everybody absolutely should know the basics of it. Um, and normally we would, we would like, I don't, there's no reason to cheat on that sort of a thing.
00:49:40
John S
Like of like, that would be be booked as equipment. So it wouldn't show up in your income statement. Um, but here's the punchline. So I stopped rambling every month. Our sales vary and our and and our cost of goods sold varies based on how much material we have bought.
00:49:56
John S
Our operating expenses are relatively smooth with the exception that usually they'll have be one or two months where we have a third payroll cycle happen in that month, which is a major major change.
00:50:00
johngrimsmo
Yep.
00:50:04
johngrimsmo
Yeah.
00:50:08
johngrimsmo
Yeah.
00:50:08
John S
So what I started doing is I look at, let's, I just did November's this morning. I look at November's revenue and then I compare November's revenue against the year to date average monthly cost of goods sold.
00:50:21
John S
So I'm smoothing those out. Cause some, some months we'll buy $40,000 worth of steel. Some of the months we buy double that. And that's not fair because we're not consuming that much steel.
00:50:32
johngrimsmo
Yeah.
00:50:32
John S
And that gets into a whole nother conversation about like what's called activity-based costing, blah, blah.
00:50:32
johngrimsmo
Mm-hmm.
00:50:36
John S
I just kind of want to know, okay, on average, our COGS should be this. And then our pay expense smoothed out as this. And that tells me what I think is a more accurate number that's not subjected to the whims of, oh, we bought a bunch of material, we had a triple payroll this month, et cetera.
00:50:54
johngrimsmo
it's good It's good to have that pulse. like I've got full-time accountant that does that. That's his job. So we get weekly reports via email, for one, slightly different. And then we have our weekly director's meeting where we go over all these numbers, production numbers, YouTube views, et cetera, et cetera.
00:51:10
John S
Oh, interesting.
00:51:10
johngrimsmo
Just kind of like point in time, scorecard type things. um And the one that I love to see the most is net cash. like
00:51:18
John S
Yeah.
00:51:18
johngrimsmo
how much cash do we have in the bank after we pay all current outstanding within 30 days bills, like obligations within 30 days sort of thing, outstanding invoices and things like that.
00:51:24
John S
Okay.
00:51:28
johngrimsmo
And yeah, you're right. Like ah last week it was, the net cash was noticeably higher than like normal.
00:51:36
John S
Oh, good.
00:51:36
johngrimsmo
And I'm like, oh sweet, what happened?
00:51:36
John S
Yeah.
00:51:37
johngrimsmo
And he goes, well, there was no payroll that week. I was like, oh, okay. It just happened to be a week that had no payroll.
00:51:42
John S
Interesting.
00:51:43
johngrimsmo
i was like, that partially explains it, but also sales were really good. So like, great.
00:51:46
John S
Yeah.
00:51:47
johngrimsmo
Okay.
00:51:48
John S
Do you, um, do you, you don't have a accounts receivable, I guess, cause you do sales.
00:51:53
johngrimsmo
Hardly. No, pretty much nothing.
00:51:54
John S
Interesting. Okay.
00:51:55
johngrimsmo
Yeah.
00:51:56
John S
Okay. Interesting. Yeah.
00:52:00
johngrimsmo
But yeah.
00:52:00
John S
Uh, that's all I got. We got, we, I feel like I'll still have a ton for next week, but we're good.
00:52:04
johngrimsmo
Yeah, exactly. It's good good to see back. Glad that you're good.
00:52:08
John S
Likewise.
00:52:08
johngrimsmo
had Had great trips. Yeah. See you next week.
00:52:12
John S
See you week.
00:52:14
johngrimsmo
right.
00:52:14
John S
Bye.
00:52:14
johngrimsmo
Bye.
00:52:15
John S
Bye.