2026-04-22 02:16:47 <KipIngram   > lisbeths: I've often thought about something really cheap and simple like that, that I could turn into "anything I could think of."  In this case I guess you'd just display the keyboard and then use the touch screen for keystrokes?
2026-04-22 02:17:23 <KipIngram   > I figure as long as I could ssh into it to do development then I could make it do anything I could imagine.  Calculator, ereader, photo viewer, etc. etc.
2026-04-22 02:17:36 <KipIngram   > All in Forth of course.  :-)
2026-04-22 03:38:45 <tpbsd       > https://dpc.pw/posts/i-dont-want-your-prs-anymore/  < short, interesting read, ai related
2026-04-22 07:05:17 <MrMobius    > calculators are tricky though because you generally don't want a touchpad on a programmable device
2026-04-22 07:05:31 <MrMobius    > so you need some kind of keypad
2026-04-22 07:47:46 <veltas      > tpbsd: I've run into that bottleneck at work too
2026-04-22 07:47:52 <veltas      > Before AI was being used even
2026-04-22 07:48:41 <veltas      > And it's not even about trust, it's just nobody really tends to actually understand the codebase you maintain as well as you do, so every innocent change turns into a crazy amount of "this is why it was actually hard but you didn't notice"
2026-04-22 07:49:20 <tpbsd       > veltas, I can see how that is a major risk
2026-04-22 07:49:25 <veltas      > And yes that reaks of technical debt, there is tech debt in my main codebase that I inherited and have been slowly but painfully paying down, while delivering on schedule
2026-04-22 07:51:50 <tpbsd       > veltas, what gets me is that nowadays so many people and industries are using AI that isnt perfect, and they dont care that a literal Sargasso Swamp of sus code is being made
2026-04-22 07:53:08 <tpbsd       > I wont use AI to make code or design anything for me because I have found too many errors in my tests. Its a magnicient Google replacement however
2026-04-22 07:54:13 <tpbsd       > so I do use AI a lot, everyday, mostly like this "[tp@fbsd15 ~]$ fabric "is there a way to convert a lepton schematic.sch file to a qucs-s.sch ?" > lepton-qucs-s-schematic-conversion-glm5.1.md"
2026-04-22 07:56:50 <veltas      > My main goal with AI is to write more automated tests
2026-04-22 07:57:13 <veltas      > Because I've always tolerated more slop in tests anyway
2026-04-22 07:57:55 <veltas      > In embedded you shouldn't have a massive core codebase anyway
2026-04-22 07:58:29 <tpbsd       > yeah, that a great idea
2026-04-22 07:59:54 <tpbsd       > Im currently working on a way to make web schematics viewable on the internet, for free, and at a professional quality for use in IRC and forums etc.
2026-04-22 08:00:55 <tpbsd       > Im sick and tired of seeing FALSTAD Java slop used on IRC electronics channels
2026-04-22 09:25:33 <veltas      > Interesting
2026-04-22 09:47:36 <tpbsd       > veltas, it's just an alternative using existing FLOSS packages
2026-04-22 09:49:16 <tpbsd       > I've currently found that Qucs-s outputting a SVG picture and saving it on Github works, here is my test image https://github.com/techman00172/techman/blob/main/export.svg
2026-04-22 09:54:08 <veltas      > Cool
2026-04-22 09:55:32 <tpbsd       > it's viewable just via the URL, free and Qucks uses SPICE, so all the ingredients are there for a decent electronics schematics base with static simulation
2026-04-22 09:56:25 <tpbsd       > it lacks animated wires with measles however :-)
2026-04-22 09:56:39 <veltas      > What's wrong with falstad?
2026-04-22 09:58:13 <tpbsd       > it's for beginners, it's java based, it lacks a wide component base, it's sim is not accurate, and I find it very annoying
2026-04-22 09:58:58 <tpbsd       > it was never made for the average electronics person, probably in the same way that Arduino wasn't
2026-04-22 09:59:44 <tpbsd       > Aeduino was made for people who had no clue about embedded, such as artists who wanted flashing lights on a artwork etc
2026-04-22 09:59:51 <tpbsd       > Arduino
2026-04-22 10:19:46 <veltas      > I guess it's an educational tool?
2026-04-22 10:23:44 <tpbsd       > yeah
2026-04-22 13:35:22 <KipIngram   > That's what I saw once I waded into the tools for working with ESP32.  The nice polished GUI environment is very limited in what you can actually DO - it was clearly intended to "idiot proof" the process as much as possible.  It "works" - you push the button and your code runs - but if you want to do anything that's even a little outside that simplest possible flow it's just not set up for it.  I
2026-04-22 13:35:25 <KipIngram   > quickly had to abandon it and work instead with the console oriented tools that work more like a normal compile/link flow.
2026-04-22 13:36:56 <KipIngram   > In my case I wanted to be able to run code I'd poked into RAM, and the GUI setup just wasn't able to do that.  In fact, it wasn't able to let you run the code you'd written in the normal way from RAM either - it by default ran it from flash, which is a fair bit slower.
2026-04-22 13:40:31 <KipIngram   > tpbsd: My hunch is that this explains that report you got about less than perfectly precision square waves; I think that "default simple path" just isn't set up in a way that encourages best real-time performance.  I think the thing CAN do such things, but it requires seizing greater control than the simple setup offers.  I've yet to 100% prove that, though.
2026-04-22 13:47:12 <tpbsd       > KipIngram, just ask your AI to compare falstad and spice simulators ?
2026-04-22 13:47:27 <tpbsd       > it will point out the differences
2026-04-22 13:47:44 <veltas      > https://letmegptthatforyou.com/
2026-04-22 15:23:31 <veltas      > Anyone here played with FastLZ?
2026-04-22 18:00:36 <KipIngram   > tpbsd: Oh, I didn't mean to be commenting on simulators.
2026-04-22 18:01:12 <KipIngram   > I'm quite convinced that "real SPICE" is the way to go if you want good results.  Provided you know how to use it well.  Good analog simulation is kind of a high art.
2026-04-22 19:23:30 <KipIngram   > I was mostly complaining about the lack of flexibility re: how to use the processor's resources in that GUI chain - it pretty much just worked one way and that was it.  And it wasn't "fastest."
2026-04-22 19:23:53 <KipIngram   > Fine for low-performance applications, but if you're really wanting to optimize it just fell short.