Thread 107081081

but if you've got CNC mills at your shop you need a lot fewer machinists than you would with manual mills

ok

>>107081105
what I'm saying is you're out of a job, zoom zoom

you need maintenance technicians unless you want to paralyze your chain for days

>>107081136
yeah but the mills aren't constantly down. you don't need one guy per workstation.

same shit applies to LLMs and codemonkeys.

>>107081081 (OP)
>lust provoking image...
i wanna build a cnc mill. what am in for?

>>107081157
yeah but a cnc mill actually works

>>107081163
if you toss a chunk of scrap metal into your mill without strapping it down and hit the cycle start it won't work either

>>107081158
depends on how shit of a result you're willing to tolerate.

let's just say that buying is cheaper, and note the price.

>>107081163
so the script work, you just didn't specify fine details.

>>107081174
thats not what youre seeing in my picrels though
a scraping script working with an api without a delay is like a mill without a locking mechanism at all
you just cannot expect it to work properly

>>107081123
But I'm a CNC machinist. What now boomer?

>>107081207
>refuse to use tools properly
>hurr durr the tools don't work
you're in the redundant pile I'm afraid

>>107081081 (OP)
Yeah but now you also need people to do your CAD and CAM work. The design engineers can do it, sure, but wouldn't you rather have someone with a cheaper hourly rate do the menial shit?

>>107081191
yeah but that invalidates vibecoding.
rectifying the optuput with subsequent prompts is one step above vibecoding- its proper prompting

not that its worth anything in my experience, though
not only you have to go through the code by hand
but youre correcting someone elses spaghetti

its just faster to write stuff by hand
idk maybe webshitters have more success with the chatbots
but thats not proper programming.

>>107081222
>CAD
>CAM
funny you should mention those. what used to take an office full of draftsmen is now one dude with Solidworks. who's often as not the design engineer himself.

>>107081213
cope tumor weeks, etc
your shit is worthless.

>>107081183
1 millie for a mill? cheaper then done by hand?
i find that hard to believe

>>107081183
>>107081248
cont.
i can program everything mind you
so i dont need to pay a codemonky on fiver to write my software

>>107081236
>that invalidates vibecoding.
you're not using a vibecoding setup.

>>107081248
a mill doesn't cost 1M dipshit. it's in the low 5 figures. but you'll end up paying more trying to make one of equivalent size and capacity yourself, because turns out building machines that can keep tolerances when loaded heavily is hard

>>107081158
>i wanna build a cnc mill. what am in for?
https://www.youtube.com/watch?v=uallSKJGoug

>>107081271
has fuck all to do with setups and shit
if the stuff youre working isnt common enough the chatbot shits itself, simple ass
that you dont know this simple truth tells me youre a programming analphabete
not only you didnt encounter such situations by your own, despite them being actually quite common
but you odnt even know how your tools work

>>107081266
the software isn't the hard part of a CNC mill. big, flat bits of metal that have correct dimensions are.

>>107081297
>has fuck all to do with setups and shit
you're using a fucking web chatbot, not a coding agent with access to a repository and the ability to dynamically look up anything in it using RAG.

>>107081285
>it's in the low 5 figures.
wtf? nah, ur shitting me
last time i checked you had to pay for the fucking quote
thats the kind of shit that happens when you're product is in the 6 digits or more

>>107081295
(you)

>>107081306
>still doesnt understand
like i said
your opinion is kinda moot and retarded

>>107081081 (OP)
>HAAS
Post a proper machine will you?

>>107081309
Some CNC mills cost that but most do not. Depending on how big you want it they start at like 10k where I'm at (EU) and go all the way up to 7 figures for big, specialized ones.

>>107081322
>>107081297
You are deliberately using a tool wrong to trip it up and then using that as some sort of proof. Again, enjoy an early retirement.

It's not that I don't understand programming, it's that you're hopelessly out of your depth in understanding how LLMs work and why what you're doing is like asking someone to land a 3 pointer with a blindfold on.

>>107081309
>last time i checked you had to pay for the fucking quote
>thats the kind of shit that happens when you're product is in the 6 digits or more
Depends on what you want. The lower limit is around $2-3k:
https://taigtools.com/product-category/micro-mills/cnc-micro-mills/
https://www.sherline.com/product-category/milling-machines/mills-cnc/
And there is no upper limit

>>107081330
yeah i just checked
there ARE mills to be bought even in the mid 4 digits
industrial shit does cost 100-500k

i want the industrial type
but its as much to have the cnc as a tool, as to develop the technologoy so that i can apply it to different projects
the whole idea is to develop all the tech id need to build a manufacturing line
5 axix wobbots, see and see
and the world is mine
ill innundate it with cheap robotic trash for the masses

>>107081358
>i want the industrial type
you will not be able to make one yourself. not without investing into all the shit the industrial machine manufacturers use to make said machine (which is a bunch of bigger machines, or buying things from suppliers who have bigger machines)

>>107081309
I got a used Mazak 410A and QT200MSY for 65k delivered, installed, and setup. both at the same time saved me a few grand.

>>107081334
>doesnt know of the statistical problem
stop pretending.
u lost. llms are barely satisfactory as search engines
and its nto like better alternatives dont exist

aerospace has mbd for instance

>>107081371
this IS enticing but im still gonna try to develop the tech
i need the technology more than i need the actual machine
>>107081368
sounds like a challenge
i have a track record of doing stuff logically i have no right to be able to
im still gonna try.

>>107081386
>the statistical problem
which of them, faggot? There are many. Go ahead and expound. Surely you're not just namedropping random shit that sounds cool to some undergradb& retard.

>>107081409
>developing the tech
I already told you, dumb codemonkey, the "tech" isn't the problem. CNC mills are 70s' technology. You don't need to invent anything new, or even write any code (you can get LinuxCNC and put open source firmware on motor controllers).

The problem is that physically manufacturing the thing is fucking expensive, if you want it to be accurate enough to be useful.

>>107081421
i alrdy explained it
if your shit isnt statistically significant it wont appear unless explicitly prompted for
or it wont appear at all if your model didnt have the stuff youre using in its data

were not talking some really obscure shit
this happens with c's intrinsics
like here
>>whats the fastest way to do x
>you can do a, b or c
>>what about d
>yes, d is the fastest way of doing things because...

you can scream all you want that llms arent shit at programming
but they are
theyre fucking worthless
and were barely scratching the surface here

>>107081448
>tech matters only if youre the one inventing it
>calls people dumb
an inability to self reflect is indicative of a sub par intellect, anon

>>107081462
cont
its not even proper intrincs in this example, actually
its basic control flow structures
>yeah, but the premioom model
i dont think i will pay for this. premium or not.

>>107081409
>technology
it's basically just some servos, quadrature encoders, and servo drivers. they are actually pretty simple. it's the precision, regidity, and sheer mass to dampen vibrations that make them expensive. you can get a small nema 24 axis drive for like $200, even less used.

>>107081527
>it's the precision, regidity, and sheer mass to dampen vibrations that make them expensive
>you can get a small nema 24 axis drive for like $200, even less used.
exactly.
this is the kind of knowledge i need.
i want to produce shit. but im a turbo poorfag with no financial backing when these things are concerned
to have even merely a shoadow of a chance i have to vertically integrate like my life depends on it. bc it does.

that means instead of buying machines- i have to build my own
the aim is not to get a cnc on the cheap, its to acquire the capability to create precision automation
building a cnc metal mill seems like the best workshop to learn that
as a bonus i end up with a working cnc mill

>>107081574
your basically going to need bigger machines to build a smaller machine. also what your talking about can be easily acquired from a textbook. it's not some mystical knowledge. machinery is old, slapping servos on them is still old. you could probably make a wood router, or something can can do like 0.5mm of aluminum, but nothing large.

>>107081752
hmm
so its more of a mech engi problem than an ee one
>baby steps
>cutting wood
makes sense.
i suppose a wood router *will be easier and cheaper to make than a metal mill
its still valuable experience and if i have something that can do simple milling i can use the same concepts, and components to already use in other projects i have
like hydroponics. its another thing i wanna do, and i want to automate it. like use a simple 2axis router -ish to do seeding/pollinization/maintenance, maybe even harvest

the reason im so hung up on irl experience is because theoretical knowledge and doing stuff in practice are two different things
also stuff like finding components/suppliers, and refining the design.
the end goal is to build automation systems and sell em. the supply chain side of things fits organically into actually building a prototype

do 'cheap' wood 5 axis cnc's exist?
i don't need steel/alu, or insane accuracy.
tolerance of 0.25-0.5mm is more than enough.

>>107081981
for max. 200x200x100mm objects

why go for this instead of DMLS? wouldn't it need even fewer human intervention beyond bed clearing and post-print processing?

>>107081981
>wood
wtf do you need a CNC for? just use a file or something

>>107082140
because of something called "material properties" dipshit

when you machine something that was originally forged as a single chunk of metal, microscopically it's still a single chunk of metal. when you sinter a metal it's not, there's all kinds of fuckery happening at the particle interfaces, introducing weaknesses in the material that may or may not be possible to solve by getting the process exactly right (but require a fuckton of human intervention in either case)

>>107081574
>its to acquire the capability to create precision automation
the ability to make precision automation is to have a bunch of big, heavy machines that are dialed in and have trained operators to use them. there's not knowledge you can use to sidestep the cost, the market is competitive as is.

>>107082251
except if the process has been refined to the point that it's valid for making rocket nozzles or HP turbine blades, then the notion of getting the material properties you want seems like a nonissue, since i doubt you're producing anything in a CNC that's going to be subjected to even harsher tolerances than those two components. so my question still stands.

>>107081918
maybe you should do some research first. there are pallet loaders, bar feeders, optical and laser measurers, hydraulic/pneumatic vices, tombstone layouts etc. the manufacturing scene is loaded with automation. my machine even checks the tool every tool change to ensure that it is not broken, sounds and alarm and texts my phone if it is. I'm not saying there isn't room left to innovate but it's going to be very competitive.

>>107082287
wont prevent me from trying.
and the market, competitive?
thats not how marketing works.

yes, there is competition as to who offers features vs who offers discounts
but the baseline is not the production cost, its the perceived value proposition. cf- luxury products

most machine shops dont have automation capabilities because they just dont have the know how, and the costs of acquiring it dont justify its acquisition
in the current day and age to do so one would have to do r&d.
most machine shops dont have a several millies laying around to finance a project that may or may not yield results

ill be doing r&d myself, so the only costs i have to deal with are the components.
a fraction of a fraction of what it would cost any established company to do the same.
i think the biggest hurdle for said companies is the programmatic side of things.
they just dont have 150k/y/person to splurge around
and the team is NOT gonna be motivated bc they hold no stakes in the enterprise, so its safe to assume that for the work one motivated person could do, you would need a whole team of people just clocking in their 8-5, dragging feet to enhance job security

i like my chances. i can program whatever i want, that part is a complete non issue
the actual problems i have to solve is the mech engi part and the ee. both of which are solved problems, its just a matter of learning what goes where, and how to get things on the cheap
and ofc iterate on the design to work out the rough edges

>>107082416
if its too competitive i can just my tech inhouse and cut out the majority of costs for my other projects
i meant automation in a wide sense
i have ideas to automate processes in the food industry
i have ideas to automate processes in construction
i wanna automate hydroponics

i dont need to sell my products to heavy industry. its just a small slice of the pie
for the rest of sectors, automation is still pretty much a virgin market
especially for small and medium enterprises.
ofc a mcdonalds pink slime factory is gonna be automated to hell and back
->but none of the kebab shops are.

holy maladaptive daydreaming thread

>>107082489
Problem is, Anon, the way you are talking about, it sounds not like you are the kind of person that will ever be able to do those things. You sound like a "idea man". If you would be capable of even the most basic part of accomplishing it, you would have started already.
Or maybe you are still just a child.
But to give you a starting point, learn how to precisely control a motor.

>>107082489
>but none of the kebab shops are.
Look up "Der Gerät"

>>107082970
>i cant get there so i will try to demotivate someone else to prevent em from getting there too
the only effect your post had, was to tell me about your own inadequacy, tranimeshitter

>>107083032
>it sounds not like you are the kind of person that will ever be able to do those things.
i dropped out of 10th grade, im self taught with everything so its normal i sound like a nigger.

i still write ml, still held leadership roles, still had the time to learn marketing, client management, still had a successful career in construction, up to having my own company several years back, i still managed to integrate a talent cultivator, one even postgrads failed to enter.
i was not joking when i said i have a track record of doing shit i have "no right to be able to". thats more or less the story of my life.
im actually a polymath, but ive been dealt a bad hand, so i live life on hard mode

>>107083070
yeah thats one thing. only 99 to go.
never seen it used where i live either

>>107081157
LLMs add the required workdlpad thoughbsince yoy have to berify everyitui they put out

>>107083570
if you used an LLM to write this post it would've been better quality

>>107082319
>the process has been refined
The process needs to be refined for every single part you fucking retard. And it won't work for all use cases.

God I can't stand fucking code monkeys showing up in random unrelated fields picking something straight off google search without even bothering to learn the basics.

Machining is the go-to process for low-to-medium volume metal manufacturing because it is straightforward, predictable and requires very little up-front investment. For many materials, you simply take a billet produced by whatever bulk production process is used and you can reliably get the same material properties out of it. There are millions of experienced machinists out there who know how to work your material and get whatever tolerance you're willing to pay for right out of the gate. Do you know what tolerances are? Because I don't think you do or you wouldn't be asking these retarded questions. Meanwhile for complex thermal processes like sintering you have to be really anal about how you do it, there aren't many people who understand the process well, and even then since the physics involved are so much higher energy and so much more complicated you will need to prototype and iterate to get anything decent out of it. And since you will be using powdered materials that get heated and cooled to join them together there are all kinds of constraints on the grain size and heat treatment that machining just doesn't have.

>>107083988
I know a high-end 3D printing place (multi million $ printers) that CNC mills parts after printing to get the surface finish and tolerances needed.
But don't expect from the tech-bro children on /g/ to understand anything abut industreal manufacturing.

>>107084679
nta, but
>And since you will be using powdered materials that get heated and cooled to join them together there are all kinds of constraints on the grain size and heat treatment that machining just doesn't have.
tolerances are only part of the the story
the crystalline structure of the resulting steel, and its adjuvants are another

>just 3dprint it
youre sounding more like a techbro than i did
and im the dude who wants to automate the whole world one kebap shop at a time, kek

>>107084890
>the crystalline structure of the resulting steel, and its adjuvants are another
I literally mentioned it multiple times in that post. And dimensional tolerances are not the only type of tolerance.
>>107084679
>sinter part
>still need to mill it to get within dimensional tolerance (no mention about material properties)
>hurr durr why don't they just sinter everything
I rest my case.

>>107085199
>I literally mentioned it multiple times in that post.
yeah... the other anon didnt catch that...
thast why i quoted the part where you were mentioning it explicitly :)


>And dimensional tolerances are not the only type of tolerance.
what are other tolerances then, in this context?
integrity of the material? i heard something to the effect of 3d printed shit having micro cavities

>>107081191
dumb fuck faggots using ai are clueless to what the fine details even are, so what's the point?

>>107081328
ugh.. kids these days..

>>107081328
these are the machines with the GPS lockout. You can't move them without the manufacturer's permission.

"cnc" reminds me of my ex who would keep making references to it, trying to see if i would be into it. and i would have been into it but i was (and am) mentally ill and wouldnt really be able to find the right foray into it. there was a few sexual dynamics there that went untapped because of me being neurotic and insane. apathy was hot though. but we didnt fuck enough at all

>>107084890
the fuel injectors for jet engines are printed, but then have a fancy annealing step to get rid of the internal stresses
but they're a part that you can't mill due to the internal channels (well, you could mill the outside, but that's mostly not critical)

Maintenance departments for all kinds of factories will still have old NC lathes for mundane shit like crankshafts and adapter plates. And mechanics that know how to do other shit than just operate a lathe.

>>107081081 (OP)
ok haas