I wish I had a more powerful computer. Mine is not set up to do any sort of AI because I have no interest in gaming. So I've never needed a fast video card. They are anyways too damn expensive for me to buy one of any substance. Maybe if those guys who fiogured a way to refactor the neural net into a more effecient format, suing way less power start releasing models that are like 80b but only use the resources of a 8b then I could them.

>>33008 I don't think you need to run a big model. These might be more useful as an assistant anyways. You can always use some online service if you only need a bigger model from time to time. And for your card 7B models should work, especially quantized ones. This should be in the LLM and chatbot thread >>250 or just search for it.

I found a really good reference book called, "Mechanisms and Mechanical Devices Sourcebook" It's a good rough guide to how things are done. Has some effeciency references and some tps on wys thngs are "generally" done. I think tiis is an excellent resource to kind of browse through. It might giove you some ideas oir you might could use it to see if you're ideas could notbe improved with a few samll changes. I foud it on the excellent free Tor bookstore, Just Another Library http://libraryfyuybp7oyidyya3ah5xvwgyx6weauoini7zyz555litmmumad.onion/ The book http://libraryfyuybp7oyidyya3ah5xvwgyx6weauoini7zyz555litmmumad.onion/meta?fname=archives/books/nonfiction/9780071704410.McGraw-Hill.Mechanisms_and_Mechanical_Devices_Sourcebook%2C_5th_Edition.Aug.2011.pdf You owe it to yourself to look at this book depository. It's great and one of the really super good features is when you load a book below it will be a selection of books that are similar in subject, though this function is sometimes way off. I would say that you could easily find enough good books on the subject of making waifus, drawing, engineering, that this one site and a lot of downloading would cover most everything.

OMG sp? My fumble fingers can not hit the right keys, and I forgot to scroll up and correct.

>>33446 Thanks for the book. I did a skim read through the first quarter, and it seems to explain each concept pretty clearly and without too much industry esoterica or jargon. I'll give it a proper read once I dive back into concept drafting. I'm chaging Orchid's plan to include demultiplexers since I want to reduce the internal frame size and motor count, so I'll need all the help I can get figuring out how to make that happen. Bookmarked the library too, the 'most downloaded' section gave me a good chuckle.

>>33453 >Demux Picrel is a great conceptual reference for a mechanical demultiplexer. Would add sensors for indexing, also rounding the mating features would help. https://www.printables.com/model/163226-mechanical-multiplexer https://www.youtube.com/watch?v=8DWK3zm9SMs

An interesting idea that could be useful Making a 3D printer into a 2D plotter with G-code Generating Gcode from normal Images https://www.reddit.com/r/ender3/comments/1f6b75a/update_generating_gcode_from_normal_images/?chainedPosts=t3_15i2rr0%2Ct3_15xq6j7 https://github.com/IridiumIO/Polycut I could see this being useful for making stencils. I have this idea of making skin from NASA's flexible armor. To get the best results you need high resolution. You could print a stencil with ink, place over a plate coated with 3D resin and then put in the sun to cure the uncovered portions, building structures by layer. Slow to do but cheap because you don't need two machines.

I was looking into Joule thief circuits and came across a site where a guy claimed he somehow put out 8x as much power as the input with a few small modifications to the circuit: https://www.homemade-circuits.com/8x-overunity-circuit-using-joule-thief/ I haven't tested it yet myself, but I'd be happy if it was just a more efficient version of the regular circuit.

A website to help you develop your own cycloid actuators. https://mevirtuoso.com/cycloidal-drive/

>>33836 Very cool, thanks Light! BTW, if I haven't greeted you before, then welcome! Cheers. :^)

>>33836 Very nice. Thanks.

>>33836 Thanks, but the big deal for me was finding the formula for OpenScad by Dan Fekete on Youtube mentioned here >>24401 Maybe using something like this her and then importing the result into a program for further modification might also work, I could imagine.

Electromagnets work via the physics of electric fields having magnetic fields. (Measured in teslas) This magnetic field can be intuitively modeled through the right hand rule. A coiled conductor, often insulated copper wire, will have a magnetic field where if curled fingers match the direction of current flow, your up pointing thumb will point in the direction of the north pole. This magnet is likely to be weak due to the magnetic field freely circulated around the wire. To have an effectively stronger electromagnet, you need to constrain the field. This requires a permeable material, a ferromagnetic material, to act as a core. This core (usually an iron based alloy) has tiny magnetic regions that are disordered, cancelling themselves out. When a field interacts with these magnetic regions, they align to that field. More current will create a stronger field around the coil, causing more of these regions to become aligned with the field. Eventually, this will reach a saturation point, where further increases in current no longer meaningfully effect the usable magnetism. After current is removed, most of these magnetic domains will revert to their disordered state. Some of these domains will remain aligned, an effect called hysteresis, leaving the core a weak permanent magnet. This is unlikely to matter for us. If it does become a problem, you can degauss it by rapidly alternating current in the coil. When designing your electromagnet (or deciding which one to use), you need to think of the magnetic circuit you need to achieve your goal. Think of what you want, then consider how to use the flow of magnetic fields to achieve that goal. Consider a voice coil actuator, you want it to smoothly, quietly, and efficiently move a load linearly. To accomplish this, you could wrap a coil of wire onto a tube of electrical steel (used for its high permeability and low hysteresis) with a neodymium magnet (high magnetic force relative to cost) in the center to move relative to the field. A spring would be added so that it would return to a resting state without requiring power. (Voice coils can have either the magnet or coil move. Usually the coil moves in applications where moving the load fast matters most, ie; speakers. Where the load must be heavier and controlled with greater stability, the magnet is moved, ie; gas control lines.) It is important to always remember, the fields weaken with distance. You want your electromagnetic fields as close to the load as possible. This is why the space between a motors stator and rotor are tiny. Why the coil wraps around the magnet as closely as possible in speakers, etc..

>>35055 POTD Let's all do SCIENCE! to it!! :D

>>35055 Thank you kiwi im planning on using a mix of iron powder and polyester based on this video https://m.youtube.com/watch?v=YQNW4ZTsg2w he said polyeutherane but i assume polyester should work. Maybe i should finish the electromagnetic robot eyes. The steel screw magnet atleast was very innefficient however since it was very sensitive to resistance and drained the batteries in minutes. Okay yes ill finish the robot eyes might as well. i did this pretty rushed. not sure if i got the ratios right but i want to do this quickly.

>>35061 too quickly..

>>35062 What are we looking at here? Did something break, Anon? Regardless, just keep moving forward, peteblank. Persistence overcomes more obstacles in life than anything else!! More than talent, more than resources, more than genius. Cheers. :^)

I'd made this as a shitpost, but I was curious how viable it might be. The biggest engineering issue I think is the arms. Even the arm joints were very basic I'd be afraid of getting the pressure calibration correct. I joke about harm to the user if the safety settings are overridden but it probably wouldn't be a good look for the product if it regularly crushed their user to death.

>>35067 ive seen these on farms, humorously called crushers, doesnt look like a person but effectively functions the same, it just hugs/squeezes the cow for a while to reduce stress, they probably use a hydrophilic piston with a pressure valve so its not possible to go beyond the designed force

>>35067 >I'd made this as a shitpost, but I was curious how viable it might be. Great idea, Anon. I think it's entirely viable! It could be a big hit if you pull it off correctly. BTW, I'd suggest you have a look at our Kiwigril thread : ( >>104 ). Looking forward to seeing your project's progress here, Anon! Cheers. :^)

>>35067 sounds familiar ;) The engineer in me has solved the human squish problem: arms driven by a winch and use elastic cord as cable to pull the arms in. Similar to one of those hands made out of plastic drinking straws. Tighter the winch winds, the tighter the hug. But its bungee cord so easy to free if something goes wrong.

>>35063 tnak you chobitsu. its my 3rd attempt. i printed a mini spoon and made sure the ratios were right this time but i think itll get stuck on the mold like the second time. i hate casting. if you want to know what im trying to do watch the youtube link i posted.

>>35071 this is not what i had in mind. i could put some tin foil paper in between the mold but its very powdery. Hardly worth the hassle. why isnt there iron rods for sale ff. i already got the materials so i might as well keep trying it to do it this way i guess. He used polyutherane though im using polyester. Very strong odor.

>>35069 Stylistically, I can appreicate it but mechanically I don't like it. Kiwi project looks like a body pillow with legs. Also, I've seen what happens if you let your guard down around a kiwi girl...

>>35084 >Also, I've seen what happens if you let your guard down around a kiwi girl... Kek. Mean as. :D Anyway, good luck Anon! Keep us all up to date with your project's progress here please. Cheers.

>>35074 Im going to have to make a bigger mold i think but using tinfoil so it wont get stuck to the mold helped.

Using waste robowaifu body-heat to power a robowaifu walking light show? https://www.youtube.com/watch?v=uNGNP5EXNWs

New research into low power AI on the edge. Here's to hoping it goes somewhere that helps us. https://arxiv.org/pdf/2502.10089

>>37139 POTD Nice find, Anon! This paper's objectives really line up with our engineering design needs here on /robowaifu/ . Cheers, Licht. :^)

>>37139 Very nice stuff you've been digging up. Thanks.

Interesting gear train technique, relying on surface friction rather than gears meshing. I suppose you could call them atomic gears given the fundamental forces governing the behaviour. IMSystems uses this technique for planetary gear systems they refer to as Archimedes Drive. It's essentially turning a bearing into a drive. Relying on rolling friction is potentially efficient, with an added bonus of being as simple as possible. It's just cylinders rolling about each other. This method also benefits from slipping. The cylinders can slide along rather than breaking gear teeth, or worse, her owners teeth! https://www.youtube.com/watch?v=z_FdCGIzG-w

>>37277 Wow, your >pics-related are really interesting, Anon. Know of any thingiverse, etc., printable resources for this concept right now? Cheers.

Any'non know about these stair-climbing wheelchair designs? >related pic: https://trashchan.xyz/robowaifu/thread/26.html#52 >(cf. threads-related : >>2983, >>24744 )

3D printable animatronic eyes https://www.xda-developers.com/3d-printed-animatronic-eyes/

>>37277 Thanks, I think I remember a video from a few years ago about it. This strikes me as something that might hold up for a long time.

Tracing designs can be a fun way to learn and experiment.

>>38254 Interesting. I like where you're going with this, Kiwi. Cheers. :^)

Compiled anthropometric data on western woman. This can serve as useful data for designing a machine which is compatible with everything an adult woman is. Presented as a single sheet to print for reference while designing.

>>38450 POTD Nice work, lad. Cheers. :^)

How to make roller teeth gears. Potentially a viable way to produce a high efficiency gearbox at home. https://www.youtube.com/watch?v=VXcuryyRGbo&t=221s

>>38529 Wow, those look amazing, Kiwi. Nice find & thanks! Cheers. :^)

>>38450 Thanks, Really helpful.

>>38450 Nice! Thank you

>>38529 Another great link. Super interesting. I'll have to really lo at it to fully understand what he's done there. Complicated.

Another great video on how brushless motors work. https://www.youtube.com/watch?v=RUS7GGv-UJs

>>38547 THE LEGO FUN HAS BEEN DOUBLED!! This is great stuff, Kiwi. If we ever teach schoolboys these things *, I'd deffo recommend this type of approach. >11'000 RPM Lego Motors!111 Michael Faraday would be so proud! :D Excellent find Anon, thanks. Cheers. :^) --- * Maybe hold yearly "Robotics Week" competitions at the robowaifu factory(s); for private boy's schools during their Summer breaks? Something similar to what NASA does (or did...not sure now)? <---> If we ever actually do go this route, then a 'Robot Design & Engineering Class' curriculum suited to students should be developed by us prior-to. We should make multi-tiered (based on the number of parts, manuals, tools, instruments, &tc., included) classroom kits available for schools to purchase; which teachers can then use during the preceding school year -- so that the boys would have their team's robots all assembled & ready to compete during the big week at the factory. Interested & qualified factory staff could be given free time during the school year, to engage remotely & regularly with the teachers & classes ahead-of-time, as consultants to help guide their team's projects to success. Trophies & other prizes for the winning school's teams in various categories. Rando giveaways. Special rates & accommodations for qualified Boy's Homes. Big wrapup party+concert. Swag for everyone. In a word: treat these kids like kings during that week! :^) <---> I do not want 'Terminatorz-R-Us' -type sh*te to be within the yearly contest themes!! Rather, let's make it about navigating Anon's flat (or the outdoors); doing the dishes by hand properly (or laundry, &tc.); recognizing/responding properly to rapidly-changing (surprising) environments; timed obstacle-course events; &tc., &tc. >tl;dr Similar to all the nominal stuff we here are already aiming to accomplish. We can also throw in lots of AI-related and other things for their robots to demonstrate. Stuff like eloquent TTS<->STT, artistry, logic-puzzles, animal-care, gardening, monitoring & cataloging plantlife & wildlife, weaving & crafts, music-making (NAME MY BAND!! :DD, &tc., to flavor things up a bit. Kind of a Country Fair type scenario such as our own @Mechnomancer (and myself; possibly others here) might relish. Simply be creative ourselves, and always take recommendations from the schoolboys, the parents, & the schools for suggested improvements to the competitions. Everything recorded & edited together nicely after the fact -- with good soundtracks -- to serve as propaganda for the next year's big event! :D <---> We can entice the schools & parents towards the project by us thereafter helping particularly-talented young men network to get placements in realworld jobs (not least of which could be our own firm's jobs!) Ofc, other interested companies can come alongside as Platinum-tier, Gold-tier, &tc. corporate sponsors (no (((military))) horsesh*te, ofc)...once the event took off. We could set up a free jobs fair for the lads right outside the competition arena venue. Various qualified Unis, as well could engage there. >ttl;dr New blood is how you keep a movement alive. TWAGMI

>>38547 >>38558 > (lego -related : >>38416 )