Episode 129

129 - Coffee, Cobots, and Coding

Jem & Justin chat Moccamasters, bad millennial design, and coffee snobbery. They cover AirShop updates, YouTube milestones, and a barrel polishing query. Plus, Vention's new sanding cobot, post-processor woes, Cursor coding tips, and the joys of TextExpander. Laura's back on payroll, profit distribution fixed.    

Watch on Youtube

DISCUSSED:

✍️ Comment or Suggest a Topic - New voice message option

  • Moccamaster makes better coffee than I used to
  • Hello Linda
  • Thanks for all you that inquired about AirShop
  • Shandy Vention Sanding Robot Turnkey - website
  • 2025 PF spready upgrade ꘎
  • New income goal, new accounts, new fixed distribution amounts, 11 months in. ꘎
  • Added Laura back on the team officially
  • Jonny post-processor gold star!
  • Getting cursor chat to talk to cursor composer? Help
  • Terrible LLM emails this week
  • New prompt tuning to compensate, still crap.
  • PS2 rides shotgun ꘎
  • Slow steppers ꘎
  • PS2 to States in – Tuesday, 13 May 2025
  • Barrel tumbling whelp ꘎


Chapters

  • 00:00 Jazzy Hands
  • 03:19 Mother-in-law Coffee Experts
  • 07:54 Launching New Ventures and Community Engagement
  • 10:51 YouTube Success and Content Creation
  • 16:34 Automation in Manufacturing and CNC Challenges
  • 22:22 AI Post-maker
  • 27:00 Navigating Post-Issue Challenges
  • 28:07 Understanding Cursor: Chat vs. Composer
  • 30:03 The Importance of Context in Coding
  • 32:03 Dictation and Its Impact on Productivity
  • 34:48 Degradation of AI Models: A Recent Concern
  • 38:01 Managing Memory in AI Agents
  • 39:53 Setting Up a Local Coding Environment
  • 43:00 Progress on the PS2 Project
  • 44:33 Timeline for Production and Shipping
  • 48:35 Spreadsheet Management and Payroll Insights
  • 51:25 Elevator Ding.mp3

---

Profit First Playlist

Classic Episodes Playlist

---

SUPPORT THE SHOW

Discuss on Show Subreddit

Show Info

Please note: Show notes contains affiliate links.

HOSTS

Jem Freeman

Castlemaine, Victoria, Australia

Like Butter | Instagram | More Links


Justin Brouillette

Portland, Oregon, USA

PDX CNC | Instagram | More Links

Transcript
Speaker:

Three, two, Jazz hands.

2

:

Woo!

3

:

I that

4

:

I guess I'm trying the desktop version today punishment.

5

:

Did you get the session back that you lost the timeline of?

6

:

Yes, you know what it was.

7

:

What?

8

:

the overlaid video of Gimble's intro loader.

9

:

I dropped a video in off his Instagram, of Michael's Instagram, which was just cool for

people watching the YouTube video, and I thought it would be easy.

10

:

broke it.

11

:

Yes.

12

:

Anyway, it going for you?

13

:

I am...

14

:

well, thank you.

15

:

I'm still waking up.

16

:

Got my coffee.

17

:

I sent you the picture of the coffee filling directly in the Mocker Master.

18

:

No way.

19

:

And what I didn't send you was the photo of it overflowing because I didn't get back to it

in time and it was just like...

20

:

spilling forth.

21

:

Which made it possibly an even better photo, but there you go.

22

:

I do love the orange one.

23

:

Yeah, the orange is good.

24

:

What color's yours?

25

:

just boring, like slate gray for the house, because that's, just, millennials.

26

:

And apparently, did you hear this?

27

:

That like Gen Z starting to see our like millennial minimalism as like what we think is

like 90s oak houses.

28

:

It's starting to look kind of like dated and like, well, it probably has for a while to

them, but.

29

:

It's just looking kind of rough, I guess.

30

:

And I'm like, no, that means we're really old.

31

:

Black doesn't get dated.

32

:

can just, black's forever.

33

:

yep.

34

:

Yep, I wear black uniform.

35

:

Do you wear a uniform?

36

:

I don't really have a sense of what you wear other than like the third of your body.

37

:

Yeah, I wanna see your pants.

38

:

Basically, yeah, I have like a couple pair of jeans.

39

:

One of, I got into long sleeve shirts last year and wore one of five of those.

40

:

Yeah, there was a while where you were wearing, I think during your winter, you were

wearing your maroon hoodie and like every one of our thumbnails.

41

:

Yeah.

42

:

Yeah.

43

:

it's just I just noticed it.

44

:

It's like They might think we're just repeating these

45

:

got about five of those Like Butter hoodies, all in various states of disrepair.

46

:

And I keep one clean one in the photo booth, just hanging on a hook that just lives up

there permanently so I can always just drag on the same clean one.

47

:

But yeah, I pretty much wear Like Butter t-shirts or hoodies all year round, everywhere,

in every context, because it's the only clean clothes I have.

48

:

It's the only clean clothes.

49

:

Yeah, well, yeah, right.

50

:

I know.

51

:

Well, clean to a point, right?

52

:

Like that's how mine are.

53

:

Workshop clean.

54

:

You don't wear these on the couch at home.

55

:

yeah, I do.

56

:

But maybe that's a reflection on our couch.

57

:

or your clean workshop.

58

:

maybe i don't know got lots of nice little notes here

59

:

let's start out by talking about how your Moccamaster I think you got me onto it and then

we bought one when we had my son because I used to make a Chemex pour over like every day,

60

:

which is what happens before you have kids.

61

:

I knew as soon as that was coming up, I was like, yeah, there's no more time to spend 10

to 15 minutes swirling coffee with a little thing.

62

:

And so I got a Mocha Master and I legitimately think it's just better than how I used to

make coffee, which is just kind of wild.

63

:

I would have never said that before about any other drip machine.

64

:

I don't understand what is so much better.

65

:

Cause it's just very basic, like of a machine.

66

:

See, I grew out of my coffee snubbery well before having children, I think.

67

:

So I don't care much anymore.

68

:

And I've had the Mockermaster a long time now, but it was back in Kensington days when I

bought it because I thought it would help keep the team on site and not wandering up the

69

:

street for takeaways.

70

:

And it did.

71

:

It did help.

72

:

You did the Google thing where you like brought in dry cleaning and childcare so they

don't leave.

73

:

That's right.

74

:

And yeah, I stopped caring about coffee a long time ago and I kind of gaslit my

mother-in-law who I, we kind of live with our mother-in-law, she's through the wall in

75

:

like a separate house and like back in the day.

76

:

when you say that every time.

77

:

She...

78

:

you know, I was very coffee snobby and I sort of trained her on how much of a coffee

snobbo was and then, you over the years I just cared less and less and less and now, like,

79

:

she's very perplexed by the fact that I'll drink anything, including instant coffee.

80

:

She'll come in asking for techniques of, how to use her aero press better.

81

:

I'm like, I don't know.

82

:

I don't care.

83

:

Do what you want.

84

:

a barista before I met her at a little coffee shop.

85

:

like she just recently, I always tried to get her to like help me make better coffee at

home.

86

:

And she just would be like, I don't care.

87

:

I don't want to do that anymore.

88

:

I don't care.

89

:

And then just recently she saw me like making on our little like crappy home espresso

machine.

90

:

She's like, you know, if you do this, it's it'll be better.

91

:

I'm like, you've waited 15 years to tell me how to make better coffee.

92

:

Come on.

93

:

Yeah, anyway, that's funny.

94

:

our machinist, because we met him.

95

:

He was our local barista in the cafe.

96

:

That's how we know him.

97

:

And then he came and worked for us and became our lead machinist over the years.

98

:

he's like, yeah, he cares the least of all of us in the shop, I think.

99

:

Like, it's like when he makes coffee in the morning, he leaves coffee grind on the bottom

of the filter and just throws it in, leaves the coffee bag open, doesn't care about stale

100

:

beans.

101

:

And I'm like, come on,

102

:

You are the barista!

103

:

It's been burnt out of him.

104

:

Yeah, I get that.

105

:

it's funny you brought up the mother-in-law thing, because I believe my mother-in-law

listens to this most often.

106

:

Hello, Linda.

107

:

And she comments on our conversations.

108

:

And I'm always like, must enjoy this, because you're way too nice otherwise to listen to

us babble about things that are pretty off in the weeds from what you're interested in.

109

:

Hello, London.

110

:

But she has also really upgraded her coffee game, probably because I come there and then

I'm fussy and she's nicely upgraded her grinder and her machines and now she's like

111

:

getting into it too.

112

:

coffee brings people together,

113

:

I'll take a milky one please Linda.

114

:

If you're right over.

115

:

it's been a little bit hectic in a new way.

116

:

I had a little bit of that experience of launching a new thing, which is it's it's that

like fun.

117

:

Crazy many more messages.

118

:

have another Instagram account now.

119

:

I think I'm on the five or six.

120

:

So a bunch of you actually are just catching the show from last week as we record this.

121

:

and that listen and we've had a bunch of people sign up.

122

:

So the response has been really good so far to our Air Shop announcement and a bunch of

people you know, not to my surprise, but just I'm happy to see that a bunch of people want

123

:

like demos and, and I'm like, yeah, great.

124

:

that's not quite ready yet, but like, you know, I'd love to show you really soon.

125

:

Hopefully within.

126

:

I'm thinking probably a week or two, probably more like two weeks.

127

:

We'll have kind of a couple more things wrapped up and

128

:

and better to show rather than, or just, just some stuff is tough sometimes to like walk

people through you know, I don't want to get, give the wrong impression of status, I

129

:

guess, because of some weird little bugs.

130

:

Don't be too precious either though, like...

131

:

Bring people on the...

132

:

I know, I know.

133

:

Bring people on the journey.

134

:

Right?

135

:

Yeah.

136

:

Yeah, I'm very, I.

137

:

Yeah, Brady said that when I was talking about demos.

138

:

I was like, I need to know what to avoid.

139

:

If it doesn't work well in demos.

140

:

And he's like, yeah, demos are always just like, you can only go to this 7-Eleven and open

the app in this circumstance.

141

:

Otherwise it's kind of implode.

142

:

I don't know, it's really fun and there's been a lot of confirmation of just basically

that our concept is what people want right now.

143

:

Like a lot of listeners here and otherwise that are like, I have been looking for a thing

that integrates those two features, inventory and quotes, and I'm excited to see what you

144

:

have.

145

:

You it's like, unless they were throwing money at you, just like I will pay anything.

146

:

It's not going to get much better than that at this point.

147

:

Yeah, yeah, it's great.

148

:

Yeah, I was chatting to one of our clients who's an upholsterer and he, I think he must

have got some glanced Airtable when he was here last and then he texted me saying what we

149

:

use and I explained it and then I was like, know, use Airtable and Quotient and bloody

blah.

150

:

You know, he's a client so he's seen our Quotient outputs.

151

:

And then I mentioned what you're doing and send him the signup form and I think, you know,

152

:

it clicked with him immediately seeing just your landing page and what you're trying to

do.

153

:

like, yeah, that's what I want.

154

:

Cool, great.

155

:

Sign up.

156

:

Right.

157

:

Yeah.

158

:

Amazing.

159

:

Yeah.

160

:

You've pushed a few people.

161

:

It's been nice.

162

:

I'm excited.

163

:

There's also just this really nice feeling.

164

:

I've said this before.

165

:

I feel like, especially because of this podcast and just talking to you that I have so

many just good people.

166

:

I guess I would call them friends.

167

:

I don't know if they would call me a friend, But just so many people that are like.

168

:

just so positively encouraging and just kind of along for anything, right?

169

:

And especially a lot of Australians that have commented and it's just a nice feeling like

to know people in a very different part of the world that you've never met.

170

:

Yeah, yeah, I feel the same.

171

:

Yeah.

172

:

Anyway, won't turn into an air shop advertising podcast.

173

:

I promise.

174

:

Just excited for the first little announcement.

175

:

Yeah, awesome.

176

:

Awesome.

177

:

How's your 55 minute YouTube video going?

178

:

decent performance?

179

:

Slowed down.

180

:

It's second to my plywood video in terms of last 10 performance, which is good.

181

:

We did also pass 10,000 subscribers basically because of that video on YouTube.

182

:

So that's kind of a nice weird little marker.

183

:

Get to another digit.

184

:

Do you get any revenue from YouTube?

185

:

Currently in the last 28 days, my estimated revenue is $101.

186

:

So.

187

:

this is not a moneymaking venture.

188

:

This is for fun or, uh, you know, genuinely still enjoy making video like that.

189

:

Like it's like a different, it scratches a different itch of creativity and like,

190

:

workflow and I was excited to use that new little gimbal camera.

191

:

the little Osmo thing was really fun.

192

:

then, yeah, I'm a tech, I'm a weird technology geek.

193

:

So it's like, just, I want to try the new thing and how does it edit it?

194

:

And is the audio good?

195

:

And yeah.

196

:

Did you watch any of it?

197

:

I haven't watched it in full yet.

198

:

I've only flicked through a little bit of it.

199

:

Bits and bobs.

200

:

Haven't had a chance.

201

:

You know, my schedule's just so full of ping pong videos after hours.

202

:

I find it hard to get anything else done.

203

:

Someone did show, sent me a YouTube video.

204

:

Another video I haven't watched, but have looked at in part.

205

:

of like, some historical manufacturing footage of a big tumbling setup, like barrel

tumbling, barrel finishing, which is cool to see.

206

:

And pretty much looked very, very similar to what we're doing actually with our new barrel

finishing with like lumps of wax going in, but it looked like they're putting, I need to

207

:

watch it in more detail.

208

:

It's in German too, which doesn't help.

209

:

But it almost looked like they were putting river stones into the steel drum and then

taking those out and replacing them with rags.

210

:

Maybe the river stones will wax.

211

:

I'm not sure.

212

:

Anyway, the rag step was interesting because it was, you know, doing a secondary process

where it's like rubbing, rubbing up and smoothing and polishing, I guess.

213

:

But I kind of, wanted to ask our audience here whether the

214

:

got any insights into barrel polishing, guess would be the best description because it's,

we're just getting a really waxy.

215

:

It's working and it's okay, but we're getting quite like a waxy finish.

216

:

Like all it's really doing is depositing wax on all the surfaces of the parts, which is

cool.

217

:

And then I put in some like Ortex or like Ecopanel like industrial P-E-T felt.

218

:

strips and just chuck them in as well thinking like they'll absorb some of the excess wax,

they'll give like a soft surface for the parts to rub against and maybe get a like a

219

:

smoother finish which they did but then they just kind of soak up wax and get fully like

impregnated and kind of stop being useful I think.

220

:

I was trying to think like what could I add into this barrel to help with the actual

polishing process smoothing and polishing

221

:

I'm a bit stuck as to what media to use and wondering whether anyone had any insights into

that.

222

:

I asked the robots and they suggested like acetyl balls, delrin beads, which I don't know.

223

:

like kind of bash it up a little bit.

224

:

Maybe, I don't know.

225

:

Maybe.

226

:

What about, maybe this makes no sense.

227

:

I don't have any much of any experience with this, but like, what about like those little

like no grit Brillo pads?

228

:

Which one's are they?

229

:

Let's preload.

230

:

Like usually they're white here, like the stuff we use for Osmo.

231

:

scratch-brite pad, similar?

232

:

Yeah, yeah, that's an interesting idea.

233

:

Got some of those.

234

:

No, but they are lightly...

235

:

I mean, that's what you'd use for polishing in some applications, Yeah, or maybe if the

inside surface of the barrel could be sort of covered with a surface like that too,

236

:

potentially, because it's really loud.

237

:

Like it's this hexagonal timber barrel.

238

:

GONG GONG GONG

239

:

dig dig dig dig dig dig dig dig dig dig dig dig dig dig dig dig dig dig dig dig dig dig

dig dig dig dig

240

:

I was trying to imagine what you would do in like an industrial setting.

241

:

You know, honestly, we should run this by David because part of what we talked about in

the tour and like what they do a lot for people, But you might have an idea of like, just

242

:

often he runs into these situations for big manufacturers where they're like, well, we

need to, we need to.

243

:

know, tumble these parts differently, like how would you do it?

244

:

And so they come up with a solution for them.

245

:

And I was imagining it more like instead of it enclosed in something that they're like

running across the surface, probably like a inclined surface.

246

:

And maybe that continues or I don't know how you keep it from, what if it's like a big U

that goes back and forth of like you would take a bowling ball.

247

:

Yeah.

248

:

shine it up, but it's just automated.

249

:

Maybe a cool machine.

250

:

I mean, that's kind of what the Tumblr barrel is doing.

251

:

It's creating a sort of infinitely rolling surface-ish.

252

:

Right, true, that's true.

253

:

Just in the case that then I guess maybe they have more chance of hitting each other, but

I don't know.

254

:

do, yeah.

255

:

Ooh, made a sale.

256

:

Flashy, flashy.

257

:

Ta-da!

258

:

Tada!

259

:

Ricky get a flashing light too?

260

:

Yeah, they're all over now.

261

:

They're in the shipping area.

262

:

They're just desk.

263

:

It's funny.

264

:

It's.

265

:

Yeah, I made a little video of.

266

:

Yeah.

267

:

He's a little camera shy, so I don't I don't push it on him.

268

:

us but it was nice to see.

269

:

Hi Ricky.

270

:

Whoo.

271

:

senior Debris sent me a Instagram post for the first to see it, but there's this new,

wait, maybe we talked about it at one point, this company called Vention, V-E-N-T-I-O-N

272

:

dot I-O.

273

:

They had like a builder of automation things.

274

:

You could like drop in robots and like extrusions and like.

275

:

build a cell basically.

276

:

think we talked about it like probably two years ago.

277

:

They just announced that they came out with a collaboration with 3M, the new rapid series

sanding solution.

278

:

And it's, it's a cobot.

279

:

It looks like the progressive, I know it's a FNUC CRX30IA, which is like the worst naming

structure.

280

:

And it's like a table that you can, they keep showing it with like cabinet door fronts.

281

:

but it's a robot that will stand to these things.

282

:

And it's like a turnkey solution.

283

:

Classically, almost no information on pricing on their website, but.

284

:

If you have to ask, you can't afford it, Justin.

285

:

Yeah.

286

:

show wood shops using it, so somebody's farting it.

287

:

Unless it's like a prototype.

288

:

Yeah, I haven't looked at the platform.

289

:

I'm super curious though to just drag some robots around, see what you can do with a

little configuration tool.

290

:

But yeah, I looked at that photo you sent me.

291

:

was like, cool.

292

:

Does it just send the fronts of cabinet doors?

293

:

Because that's kind of like the easiest thing to send in my mind.

294

:

But maybe cool.

295

:

I don't know.

296

:

If it can break edges, if it can send the bevel on my kid apart shelves, then.

297

:

Maybe I'm interested.

298

:

I don't know.

299

:

They show it on the second link I have there on their website as standing a big long

plywood panel.

300

:

Looks like it's like two and a half meters long, which would be surprising to me if it

could reach all that, but it might be on a track too.

301

:

Yeah, mean, I don't know.

302

:

Curious to learn more.

303

:

It won't let me sign up with an Australian phone number, another time.

304

:

Yeah, I mean, our big wide belt sander obviously does good surface sanding.

305

:

You feed things out and it sands surfaces nicely.

306

:

What it doesn't do is, you know, do any edge breaking or anything like that.

307

:

And I've seen wide belt setups where they're a little bit more fancy and they've got like

kind of brushed stages within them that can get down into like a molded cabinet door

308

:

detail, which, you know, is really not a thing that I'd ever do, but it can, you know, get

over an edge a little bit and break a sharp edge on a front of a cabinet door or, you

309

:

know.

310

:

Mm-hmm.

311

:

get down into a V-groove and actually smooth it and things like that.

312

:

But yeah.

313

:

But you know, we've talked about here, like the bevels on our kid apart shelves are one of

the more labor-intensive sanding steps that we do.

314

:

If you could teach a robot to get around, that radius and sand at the 60 degrees, then

great.

315

:

Be cool.

316

:

Seems so doable, but I mean, there's a lot of factors to it, I suppose.

317

:

on this venture page, I mean, I'm sure they're always like overly aggressive about ROI

numbers.

318

:

They're claiming a two-year payback on investment.

319

:

I saw another one somewhere on a different page of a different setup.

320

:

I think it was the Instagram post said one year

321

:

of the ROI, is like pretty fast.

322

:

But that's, you know, it's kind of, it's, pretty appealing at that point to like, if you

had the cash or you could finance it in your case to like, you have a use case in this, in

323

:

this situation where you make a lot of those dang bevels and if it frees up somebody to do

it differently or like, I mean, I hate to say it, but I do imagine this could be something

324

:

that a robot could do better than a human, like those kinds of bevel sanding.

325

:

Yeah, that's right.

326

:

easy to go around a corner with a consistent motion, even if you've done it a million

times.

327

:

Yeah, and Ben will do it differently to how I would do it subtly.

328

:

And not that that matters that much in this particular application, but consistency is

nice.

329

:

Yeah.

330

:

could do it better, potentially.

331

:

Andy

332

:

my name for it.

333

:

shandy is just like a terrible drink that you have in a pub in Australia somewhere.

334

:

You have those?

335

:

You shandies?

336

:

Yep.

337

:

Cool.

338

:

Yeah, maybe.

339

:

I don't know.

340

:

Have a shandy to go, please Linda.

341

:

Hehehehehe

342

:

They just give you a robot.

343

:

Yes.

344

:

Fantastic.

345

:

Woohoo!

346

:

Speaking of robots,

347

:

over the years with on Trinity, which is our newer, faster, better flatbed CNC, the

Canadian built multicam.

348

:

We've had various little weird, weird ghosty bug, like ghost in the machine sort of bugs

from time to time.

349

:

Like it will start a cut and instead of driving at the intended feed rate, it will go at

like

350

:

200 millimeters a minute and just like crawl and burn the tool and then get to a corner

and go, oh, off we go now and just like weird little thing or like you cut a spiral and if

351

:

you pause in the middle of a spiral and then hit resume, it will just like go, woo, I'm

going this way now and just like cut through your work and completely the wrong direction.

352

:

And John and I have always been a bit perplexed and stuck with it.

353

:

And we've learned certain little tricks in Fusion where it's like, okay, if you add a

leading point here, a start point here, then that'll override that issue.

354

:

And it's basically post processor incompatibilities or bugs.

355

:

And that's what it comes down to.

356

:

we're cutting these jobs, cutting these parts this week for some light fittings that we're

making for a client.

357

:

And it's all circular work, like outer circle, inner diameter, circular cutouts within

that.

358

:

And he cut the first few test parts and they weren't concentric, like they weren't round.

359

:

You put two circles on top of each other and twist them 90 degrees and then like you get a

difference in edge by like a couple of millimeters, so significant.

360

:

and but then you cut exactly the same file on the other machine on camera and they come

out round and really spooky but then you cut a circle of a slightly different diameter and

361

:

it cuts perfectly so it's just like this these spooky little post processor bugs which we

can't don't know how to deal with sometimes and johnny comes into the office the other day

362

:

he's like yeah i just i made a new post processor from scratch using chat ChatGPT t

363

:

I was like, what?

364

:

You did what?

365

:

No, based off the good one.

366

:

So based off Cameron's post, because Cameron and Trinity have different line arc needs.

367

:

use the two different, you know, G-code, you've got the two different arc systems, which I

can't explain, off the top of my head.

368

:

And they use the different systems anyway.

369

:

So he's...

370

:

He fed them, he fed both posts into chatty, fed good code, bad code, and then got it to

build a new post from scratch for Trinity and he's been testing it running and so far so

371

:

good.

372

:

And I was like, shit, yeah, that's awesome.

373

:

Never occurred to me.

374

:

Those kind of things do freak me out a little bit.

375

:

The level of...

376

:

like I'd probably do it for the router.

377

:

I'd get pretty nervous about the mill just because it's like so apt to just destroy

itself.

378

:

But I've...

379

:

I mean I used to do...

380

:

before there was these LLM things I used...

381

:

I have modified both of them myself so...

382

:

I guess that might be more risky considering my coding capabilities.

383

:

I don't really know.

384

:

Yeah, I've done like little mods like, you know, little one line here, one line there sort

of stuff, but not never occurred to me to do a complete overhaul.

385

:

It's a good point.

386

:

We should do the same because we have, was going to say for the shop saver, the shop saver

fusion post has some glitch in it where if I can never remember which one it is, you know,

387

:

all those like feed rate settings on a, underneath the tool, you select the tool and tool

path.

388

:

on that first page of the settings, you get like feed rate lead in lead out.

389

:

It's like if too many of those are the same, I think it's if they match the feed rate, it

will basically glitch out.

390

:

I don't know what happens, but it's like it's like almost like two negatives make a

positive situation and it will just go at rapid speed.

391

:

It'll just start cutting it rapid like the exact same thing you're talking about.

392

:

It'll ramp in and then it just has no feed rate.

393

:

It just blasts away and it's got to be some logic about how it's

394

:

looking at when it needs to post a feed rate, you know, like when it changes, it's

thinking like, no, it already said it was this.

395

:

And so it doesn't post it for some reason.

396

:

I've never tried to fix it, but it's annoying.

397

:

The mill doesn't do that.

398

:

surely that's a post issue as well, Gotta be,

399

:

just never wanted to spend the time on it because it's really not that hard to avoid.

400

:

I get, in your case, that seems like it's way worse because ours never goes on random

tangents.

401

:

It doesn't go on joy rides.

402

:

spooky when it does.

403

:

not a fun time.

404

:

Yeah.

405

:

Don't know.

406

:

Maybe I I three actually.

407

:

Yeah, just in his own chatty account.

408

:

I think he was using O3 maybe.

409

:

I imagine cursor could do a good job of post stuff, Be a good environment to work on

posts.

410

:

had a question for you actually on cursor, given that you're, I feel like you're more of a

pro user now that you're a software developer.

411

:

I'm a little perplexed by the...

412

:

difference between the chat function and the composer function.

413

:

I've watched enough videos to know some answers

414

:

what I haven't done.

415

:

I haven't done any research.

416

:

I just come to you.

417

:

So sometimes I'll jump on new project.

418

:

Like I was trying to make a Shopify app this week and I jumped into chat and I workshopped

it in chat.

419

:

And then I was like, Oh, that's all looking good.

420

:

Now what?

421

:

And then I go into composer and it didn't seem to share the context of what I was or what

I'd already discussed.

422

:

for some reason.

423

:

Balloons!

424

:

I stopped using chat as my first answer.

425

:

was using it and I think I had just started using cursor when they didn't have composer

and like the next week or something they put in composer and it seems to be the future of

426

:

how they're going to use it where it's more, it might be more context to where it's

427

:

I'm not going to give a good answer, but basically it seems like the way you should use

composer.

428

:

That's, that's the answer.

429

:

yeah.

430

:

Yeah.

431

:

There's, there's something, some reason, but I watched, I watched the, the prime time guy,

the streamer I watch that does discussions about code things and development.

432

:

And he kind of did a whole thing on, why to use that.

433

:

think, and, I think they might get rid of chat at some point, maybe I'm not really sure,

but.

434

:

Okay.

435

:

Because like, yeah, like the distinction kind of made sense in my brain of like, I'm going

to go over here and just sort of chat through a problem, but it's not going to make direct

436

:

edits to my code.

437

:

But then having had that conversation, wanted to then go intuitively, I then wanted to go

to Composer and be like, right, let's implement that now.

438

:

And then it's like, I don't know what you're talking about.

439

:

okay.

440

:

All right, so I just asked GPT to tell me about cursor.

441

:

Chat is best used for quick AI assisted coding debugging without switching context.

442

:

You get inline explanations, code suggestions, interactive back and forth.

443

:

Composer is more structured, good for generating or refactoring larger code blocks.

444

:

It's like an AI assisted code editor where you input prompts and it produces a structured

response.

445

:

man.

446

:

I've seen it described as like, makes sense.

447

:

It's like I heard on the VergeCast, they're discussing we've gone along with this idea

that chat is the way to engage with these LLMs because it was like the UI that made sense

448

:

in the beginning, but realistically it's like probably not the right fit for a lot of

things.

449

:

You know, like it should be something else.

450

:

that chat's gonna fade away to some extent in a lot of these tools, I reckon.

451

:

Yep, Weeee!

452

:

Thank you.

453

:

I knew you'd know the answers.

454

:

Yeah.

455

:

Some.

456

:

Anyway, I didn't have any luck building my Shopify app, but I was trying to do it in a

rush.

457

:

Great.

458

:

Make changes to your Shopify structure in a rush with experimental tools and then walk

away.

459

:

Push So I need to sit down and have some consolidated time pondering because I'm excited

to have a crack at it.

460

:

It's like it's replacing a $50 a month.

461

:

in a store that kind of handles material variants and swatches and stuff.

462

:

Like, I'm sure I can work this out.

463

:

But more of an evening work weekend project, perhaps.

464

:

Yeah.

465

:

think is pretty impactful, especially when working in cursor is,

466

:

I think you're better at this than I am like I start like a too much like a chat and I

don't give enough Context in the beginning and I think that sets to off down like a

467

:

confusing trajectory sometimes for how it answers whereas like if you give it like a

pretty Detailed I want this I don't want that I want it to you know work like this I need

468

:

you to use this type of coding and make a pretty extensive first request

469

:

or like guidance on how you want to work.

470

:

think it does seem to treat that better.

471

:

it, yeah.

472

:

Cause it's always just looking for the next answer, right?

473

:

So it's like, if it has this bigger piece of data to start with, I feel like you get lost

often in the discussion of the like, no, no, I only meant it to be like this.

474

:

And then it gets confused and it's like, well, wait.

475

:

What did you want to change?

476

:

I'm going to just whack them all back and forth answers.

477

:

That's what I love so much about dictation and like, because you can give so much more of

that nuanced sort of context, you know, waffling three minutes.

478

:

DUBUDUBUDUBUDUBUDUBUDU

479

:

I should use it more and more.

480

:

I started using it so much more because of having a baby.

481

:

I mean, I've always been a big Siri AirPods user, man, the things like even sitting at my

desk trying to hold a kid.

482

:

and get a couple things done.

483

:

You just hit the dictation button and talk and keep their hands off the keyboard.

484

:

So much easier than trying to...

485

:

No, no, don't hit the button.

486

:

I know it's shiny.

487

:

yeah, I think I've been partly using it way more because I have an office to myself these

days.

488

:

That's true, that is a big difference.

489

:

I'm not in a shared office as I was a year ago and so I used to be a little bit

self-conscious about like just constantly talking to my computer or phone whereas now I'm

490

:

happy to just like waffle away and it's kind of broken my awkwardness about it like

someone can come into the office now and I'll know they see me doing it all the time so

491

:

I'll just keep chatting away doing what I do whisper whisper whisper whisper

492

:

to my computer.

493

:

No, you didn't see anything.

494

:

Someone came in.

495

:

Someone's here.

496

:

just say, did you just call that computer a name?

497

:

Like, are you referring to it as a person?

498

:

Sometimes.

499

:

Depends which agent I'm talking to.

500

:

You're just like, I'm talking to my therapist, go away.

501

:

Did you see the sign?

502

:

I don't know.

503

:

Have you noticed any degradation in models these last few weeks?

504

:

Yeah, I'm...

505

:

had a terrible time this week.

506

:

Yeah, yeah, I'm so like, I'm ready to blame it on that, but also I'm like, well, what else

changed?

507

:

know, like what else?

508

:

Like I've been trying, I was trying to use that circle analysis script in Rhino that I

brought up a while back and it just used to work fine.

509

:

And then I went to use it and I got a new error.

510

:

In the script that was like the circle, something, something doesn't work.

511

:

It's not in the library or something.

512

:

So I like, couldn't find where I originally, which, which tool helped me make it.

513

:

And so I started a new, new thing with, I think, three mini maybe in, in something.

514

:

don't remember where I was in.

515

:

And it just couldn't give me answers.

516

:

It couldn't solve the problem.

517

:

It took 30 different attempts.

518

:

of it trying to give me code to fix that one little piece and then it wouldn't find all

the actual holes again.

519

:

Like it missed the smallest ones.

520

:

And I was like, well, what good is this?

521

:

know, like I've got this partially crappy script.

522

:

And I suppose that's always the case when you're talking about these tools versus a human

that knows how to write code is they're not gonna just make a partial solution probably

523

:

that misses.

524

:

one's end of the analysis of your parts, guess.

525

:

I don't know.

526

:

Just yes, is the short answer.

527

:

Yeah, I don't know what happened this week.

528

:

just like, cause I use my email writing agent heaps many, many times a day and for, know,

for basically every email I send.

529

:

And I just noticed it like starting to really lose context.

530

:

So like it's, it's looking at my Gmail, it sees the customer email.

531

:

I dictate a response and then it was just doing really wacky stuff like

532

:

getting roles reversed of like taking something from the customer's email and then writing

it as if I was saying that.

533

:

So getting confused contextually and just putting like starting like some hallucinations,

which is pretty rare for it.

534

:

Like, no, I didn't tell you that price.

535

:

Like why are you putting pricing in email?

536

:

didn't even mention pricing.

537

:

Why would you do that?

538

:

So there's a lot of cursing.

539

:

this them theory?

540

:

Yeah, also all three Sim Theory agents, but with various different LLMs, like trying it

across different ones.

541

:

I had sent you that I had noticed that it was passing more between threads than I had

expected it to.

542

:

Like it had remembered, I don't even remember what the detail was, where I was asking a

question and one of the agents, think it was my Portland CNC agent, and it like gave me

543

:

all this context of something that I hadn't explicitly given it, right?

544

:

Like I hadn't told it to keep this in memory or whatever its thing is.

545

:

And it, shockingly, it was just like, was like, whoa, you remembered I told you that

somehow, someplace, you know, like, I was, I was surprised.

546

:

And so I wonder if that also is a...

547

:

If it degrades and is remembering things, maybe you didn't want it to, or trying to always

inject itself in a certain way.

548

:

I don't know.

549

:

Do you ever go through and filter out memories in an agent?

550

:

I should I just get I just don't want to spend the time to do it.

551

:

I think is what it ends up being Do you do it?

552

:

Yeah, I do gardening in there.

553

:

Weeding, because sometimes some of the crap that drops in is like, no, don't remember

that.

554

:

But there's some great stuff in there as well.

555

:

it would add stuff, right?

556

:

Yeah, that was cool.

557

:

That was cool because you could tell when it was remembering weird things.

558

:

But I really like the way they've built memories in SimFury.

559

:

think the way it functions is great.

560

:

But I do think it's important to go in there and do some gardening from time to time and

just cale and trim things.

561

:

But I ended up rewriting all my agent prompts for emails and stuff multiple times and...

562

:

I just ended up going back to an old version, to be honest.

563

:

I was like, okay, this works.

564

:

Let's just stick with this one for now.

565

:

Yeah, I don't know.

566

:

time on that kind of stuff lately.

567

:

Just like throwing stuff at it and wanting it to work and then running away and not trying

to like...

568

:

Well, I think I said this forever ago.

569

:

I keep having this feeling of like not wanting to invest anything extra into any of these

systems because we keep switching...

570

:

I keep switching between which one I want to use based on some new thing.

571

:

Something new has come out and it's like, actually GPT is really good now, let me use

that.

572

:

try to go back to sim theory and it's, got some new things.

573

:

So it's like, you know, it's a sunk, it's not a sunk cost.

574

:

I suppose it is a sunk cost and then I, I don't want to sink any extra time into it other

than just give me an answer.

575

:

Yeah, yeah, totally.

576

:

user needs to start over with AWS configuration.

577

:

Yeah, seems accurate.

578

:

I've just setting up my little local coding environment for the app.

579

:

It's messy.

580

:

It is one of the most complicated things that I've encountered to like...

581

:

It, it's just like six layers deep of like, have to log in and all these different ways

and do all these different commands and have all these packages installed.

582

:

then even when you get it all right, there's still issues.

583

:

And you're like, I don't really know what I did or how I got here.

584

:

You know, like, yeah, just trying to, it's trying to create a local coding environment.

585

:

So you can test stuff that isn't in the actual app.

586

:

that breaks my brain even after months of.

587

:

figuring out how to do it well and then coming back to it it's like broken and you're

like, what happened?

588

:

I don't understand.

589

:

Yeah, anyway.

590

:

The PS2 was looking good yesterday, driving around.

591

:

Rowe was here in the workshop working on it.

592

:

yeah, it drives around.

593

:

Have you seen it moving?

594

:

Did I?

595

:

I sent you the test cuts, didn't I?

596

:

Okay.

597

:

Okay.

598

:

I was thinking you meant like it was on a movable robotic base or something.

599

:

It's like driving around.

600

:

How's it driving around?

601

:

around every week actually because Rory lives in Melbourne and takes it home every week in

the front seat of their Celica.

602

:

And so I've got this photo of it just strapped in the front seat with the seatbelt on.

603

:

But no, it's looking pretty good.

604

:

But the steppers that we've put in are proving to be slow, which isn't super critical at

this point or to keep.

605

:

testing and get some proper test cuts going, but it is significantly slower than the

ClearPath driven PS1, which is not surprising.

606

:

Much cheaper hardware, but, yeah, might end up swapping them out.

607

:

is it worth, what's the cost difference just relatively?

608

:

I don't even know.

609

:

don't know, it's probably like a hundred bucks worth of steppers worth of this is 600

bucks worth of clear paths or something, I don't know.

610

:

Off the top of my head.

611

:

suppose in a production environment though that money is definitely worth it even for

small, you know, for testing it's what it is but I don't know.

612

:

Yeah.

613

:

Yeah, yeah, yeah.

614

:

We'll see.

615

:

Yeah, exactly.

616

:

speed of producing things.

617

:

Yeah, I do love the ClearPath system.

618

:

Whenever I've had to fiddle with it, I'm always impressed.

619

:

It's rad.

620

:

Yeah, the tuning's a delight and implementation's straightforward and yeah, they've just

ticked a lot of boxes, I think.

621

:

Very nice.

622

:

Yeah.

623

:

What are you working through now?

624

:

Design changes, refinement in

625

:

Kind of lots of little problems.

626

:

I'm sure you're familiar with that stage.

627

:

But generally speaking, we're at a point where we're setting up the workspace coordinates

and putting production, its first production code into it so that we can get some actual

628

:

production part test cuts going.

629

:

and then expecting that we'll find issues with that in terms of rigidity and then starting

to like one step at a time going cool let's replace this assembly with machined aluminium

630

:

parts test it again and just kind of work through that until we've got a robust tool

basically

631

:

Very cool.

632

:

So like you'll have some interface where like different parts will be kind of just ready

to like click start making end cap kind of thing.

633

:

Yeah, correct.

634

:

That's the idea.

635

:

The little CNC controller is just sort of PC desktop driven at the moment.

636

:

It's actually a web page.

637

:

You go on to its own, it's got its own Wi-Fi network.

638

:

You log onto that and you jog and control the machine and drop code onto it from there.

639

:

So we'll see.

640

:

It's all, yeah, it's all pretty rough and ready at the moment.

641

:

But ultimately, yeah, some sort of interface where you can just quickly select between

the, you know, the four different file types and hit go.

642

:

It's the aim.

643

:

Yeah, that's really cool.

644

:

Hmm.

645

:

Yeah.

646

:

is the dreaded question.

647

:

you have like a timeline some.

648

:

love to be ready to ship one to the States in three months.

649

:

Whoo!

650

:

Heard it here first.

651

:

If I write it down, it's in gold.

652

:

one.

653

:

Yep, that's the plan.

654

:

Let's see, I'll use my little script to see what that date should be.

655

:

Text Expander.

656

:

I can see you doing it.

657

:

Trying to find the right shortcut in air table.

658

:

while I'm talking, I couldn't remember the stupid command and FD.

659

:

So three months.

660

:

Tuesday, May 13th, 2025.

661

:

I've also got ones that will do like only weekdays, which are real nice when you're like

estimating timelines for clients.

662

:

You're such a TextExpander pro.

663

:

Dude, GPT just, I had so many things where I was like, my God, I wanna make this text

expander snippet.

664

:

And like, I couldn't figure out how to do a little coding to make it work.

665

:

And it just was like, yep, here you go, this is easy.

666

:

You know?

667

:

Like, it'd be like the most complicated stuff and it just, it's like nothing for it.

668

:

Do you have a cheat sheet to remember all the shortcuts?

669

:

They just, do you find yourself searching?

670

:

yeah, I use the search thing to pop up the ones I don't remember.

671

:

without it.

672

:

Yeah, me neither.

673

:

It's awesome.

674

:

I can't believe how long it took me to get onto it.

675

:

Right.

676

:

Speaking of useless tools that you got me onto...

677

:

The Raycasts?

678

:

What the hell, man?

679

:

I don't get it.

680

:

You don't like it still?

681

:

Nah, don't like it.

682

:

I thought you're gonna go, that's great.

683

:

No, this stupid.

684

:

The one thing I got excited about which I messaged you was like, it can eject all my hard

drives and USBs in one fell swoop every time I go home.

685

:

You could pick a little shortcut for it.

686

:

But it doesn't work!

687

:

Doesn't work!

688

:

Nothing happens!

689

:

shortcut and then have it run the shortcut.

690

:

No, but this shortcut doesn't do anything.

691

:

Breakasts did not reject all my disks, as it promised.

692

:

I don't want to try that right now.

693

:

It does say, just says EJX all disks.

694

:

I don't know.

695

:

I, a Mac user and I'm not using it to put files onto a drive for the mill, which literally

is the only time in my memory that it matters if you eject the disks.

696

:

It will corrupt the stupid USB stick every time or whatever, CF card.

697

:

It's the only time I've ever needed to eject this before pulling them out, but otherwise I

just yank everything out.

698

:

don't.

699

:

Yeah, I yank them all out too.

700

:

I just get worried sometimes when I've got like all my, you know, years worth of photos,

my little SSDs next to me every day.

701

:

But I'll be the to...

702

:

How've you gone on time, boss?

703

:

Showmaster?

704

:

55 minutes!

705

:

god, like it's this.

706

:

daddy.

707

:

That's not it.

708

:

I thought it was P.

709

:

god, where did it go?

710

:

here it is.

711

:

And now it's profit time.

712

:

I'm gonna have to make that P slide over.

713

:

for profit.

714

:

I looked away for a second.

715

:

I looked back and your head was huge looking for buttons.

716

:

We've finally rejigged our spreadsheet.

717

:

Made a 2025 version and Laura did all of it.

718

:

I had little involvement.

719

:

Laura went into mad spreadsheet mode last week.

720

:

this week and fixed.

721

:

We found a whole bunch of errors, interestingly.

722

:

I had somehow undercooked payroll last year in terms of what it was allocating for payroll

was under by a fair bit.

723

:

It was weird.

724

:

I don't know how I did it.

725

:

Like I had people's salaries in there correctly, but somehow in my calcs I'd stuffed it up

and also materials were way under and so like

726

:

Yeah, she's reconfigured it all so to reflect more accurately on our last 12 months of

trading.

727

:

But what she's also done is made certain accounts fixed, so like payroll.

728

:

specifically is like a fixed distribution.

729

:

We know how much payroll is every fortnight, so that's no longer percentage based.

730

:

So I was weird doing the first distribution in the new system on Monday of like, oh wow,

materials is huge and payrolls, you know, because we had such a good week of sales last

731

:

week, I was distributing like 30k in one hit, which is unusually big.

732

:

And I was like, wow, materials is like,

733

:

coming through at almost 50 % of this.

734

:

That seems high.

735

:

it's because of that sort of imbalanced some fixed distribution, some percentage based.

736

:

So it's all balancing out, but yeah, exciting.

737

:

And, you know, some regrets of not having done it myself, but awesome to have had Laura do

it.

738

:

She's very good at that stuff.

739

:

She's a closet accountant and very good at spreadsheet formulas and she enjoys it.

740

:

so, it's kind of also kind of marks the point of putting Laura back on payroll officially

in the company, which is exciting because it's been bloody, I don't know.

741

:

I don't even know how many years, could be 10 years since Laura was on payroll.

742

:

So maybe, yeah, could be back to plasma days.

743

:

Cool.

744

:

you know, she's always been heavily involved, now she's, you day a week.

745

:

Day and a half a week, yeah, she's getting paid for it again, which is good.

746

:

That's Erin's always she's always like, am going to get paid paid for working paid for

doing something when I always like teaser about like you should come come do something for

747

:

us.

748

:

Yeah.

749

:

So that's been good.

750

:

That is all really.

751

:

that, what I realized is I took your sheet and lightly modified like two pages and I have

no good nexus for like, keep track of what the distributions are, but I really need to

752

:

rework it.

753

:

And I realized that probably should just make an agent in SIM theory for this.

754

:

We have six seconds.

755

:

We gotta go!

All Episodes Previous Episode

About the Podcast

Show artwork for Parts Department
Parts Department
Justin Brouillette (Portland CNC) and Jem Freeman (Like Butter) discuss CNC machines, their product design and manufacturing businesses, and every kind of tool that they fancy.

Listen for free

About your hosts

Profile picture for Jem Freeman

Jem Freeman

Co-founder and director of Like Butter, a CNC focussed timber design and manufacturing business in their purpose-built solar-powered workshop. Castlemaine, VIC, Australia.
Profile picture for Justin Brouillette

Justin Brouillette

Founder of Portland CNC & Nack