2025-11-28 02:23:51 <MrMobius    > So as you find things that it got wrong, can you really keep telling it to fix its mistakes? I read about the window/buffer some AIs have where it starts dropping your instructions from the beginning when the buffer gets full
2025-11-28 02:24:31 <MrMobius    > isene: I'm also curious if you plan to modify the code yourself at any point
2025-11-28 03:33:52 <isene       > CC, with the setup I have, is a lot better than me at writing and fixing code. Even my prime language,  Ruby. For SimplicityOS, I will probably never touch the code myself. It's similar to when we get fully self-driving cars - humans will not be allowed to drive as it will be too risky with all the cars moving at 100km/h with 0.5m space in between.
2025-11-28 03:37:50 <KipIngram   > That's  disaster waiting to happen.  Every so often something will go wrong, and instead of a fender bender or a couple of fatalities you'll get hundreds.  Hopefully I will be long dead by the time that happens - I ENJOY driving.
2025-11-28 03:38:04 <KipIngram   > It's also a target for terrorist hacking.
2025-11-28 03:38:27 <KipIngram   > Because - do you really think we're going to suddenly figure ou thow to make our systems truly secure?  I think that ship has well and truly sailed.
2025-11-28 03:59:46 <MrMobius    > I'll have to remain skeptical then :)
2025-11-28 05:39:47 <tpbsd       > I used a AI to make a helix plugin database reader, worked fine, a few changes later I was very happy. Then Lua upgraded the version and now it doesnt work with lots of complex errors, and I have no clue. Looking back it was a waste of time, I'd have been better off learning Lua znd designing the plugin myself, perhaps using the AI as a tutor
2025-11-28 05:40:54 <tpbsd       > AI definitely has a place in the world in my opinion, but not designing code, as a helper, tutor for the clueleless perhaps
2025-11-28 07:42:14 <veltas      > That's what I'm worried about
2025-11-28 07:51:56 <tpbsd       > all my life Ive done every design by myself, Ive never had anyone to bounce ideas off etc. I like that AI can provide a kind of interface because Ive only ever had books to learn from. The Internet was helpful for a while but much of the information on it is just wrong now, ... too much cruft
2025-11-28 07:52:49 <tpbsd       > Ai can be4 equally wrong, same as a book, it's just easier in some ways to use
2025-11-28 07:53:58 <tpbsd       > I think that AI is just the same old stuff with a new face. Its not alive or sentient, it's just a program, GIGO
2025-11-28 07:56:58 <dave0       > tpbsd: technically you could make an LLM in brainfuck ... i think living things dont run on brainfuck :-)
2025-11-28 07:59:44 <tpbsd       > dave0, Im not even a programmer, I'm a hardware guy, pcbs and soldering iron, CAD, schematics, so my viewpoint is 180 degrees from programmers, usually
2025-11-28 08:01:00 <tpbsd       > and LLM's can be quite helpful to prople who have no electronics knowledge or experience, I rate them as junior technician capable
2025-11-28 08:01:59 <tpbsd       > I'm only on this ch because I use Forth for embedded, where it excells, there is literally nothing like it
2025-11-28 08:03:12 <tpbsd       > Im one of the arcane 1% you'll never meet because we are so rare and embedded is totally owned by the C language
2025-11-28 08:04:12 <tpbsd       > in general electronics people dont program and those that do are hanging on by the skin of their teeth to C and believe that it's the answer
2025-11-28 08:04:55 <tpbsd       > all they know about Forth is that it's ancient, obsolete, and useless in embedded
2025-11-28 08:05:16 <tpbsd       > and they wont be caught dead bothering with it
2025-11-28 09:18:33 <isene       > When automatic exchanges replaced human operators, there were protests and union strikes. Operators argued machines couldn't handle emergencies or provide the "human touch." Phone companies faced boycotts. Now we'd find human-switched calls absurd.
2025-11-28 09:20:42 <isene       > The Luddites weren't wrong that looms produced inferior cloth initially. The question isn't whether the technology is perfect. It's whether it improves faster than alternatives.
2025-11-28 09:44:12 <tpbsd       > isene, what about the human cost of the factory looms ? for instance the default home cottage industried had been using looms for a long time, but it was individual artisans at home in villages
2025-11-28 09:45:38 <tpbsd       > the big problem as I understand the loom issue is that the rich moved the looms into factories where they were powered by a motor and the rotary motion was transferred everywhere by belts
2025-11-28 09:46:36 <tpbsd       > then they produced lower cost material and put the artisans out of business with low cost factory labour working 9 - 5
2025-11-28 09:48:01 <tpbsd       > the loom itself wasnt the problem, thats just a tool used by artisans at home in villages. They worked at home and got 100% of their labour profit. No 9 -5, no transport to and fro work etc
2025-11-28 09:48:55 <tpbsd       > it was a family and village afair at a time when it took a woman 6 months to make one set of clothes for her family, working at home
2025-11-28 09:49:40 <tpbsd       > the 'industrial revolution' destroyed that way of life
2025-11-28 09:50:42 <veltas      > You could imagine though a world where without any force these peoples lives would have been destroyed by competition alone
2025-11-28 09:51:27 <isene       > The point is that there are plenty of historical examples (like the Red Flag Act) where people were at first sceptical to innovation and automation, then later we take it for granted. Like the ATMs, elevator operators, agricultural machinery, typesetting, the synthesizer, etc. AI just follows suit.
2025-11-28 09:52:03 <veltas      > If 'AI' really turns into AI then it won't be like any of the previous changes
2025-11-28 09:52:26 <tpbsd       > veltas, from what Ive read, artisans made the clothinng which was of high quality and expensive and people were used to that
2025-11-28 09:52:34 <isene       > And that is true. That is the main thing I'm covering in my book.
2025-11-28 09:52:59 <isene       > The real underlying difference between humans and AI...
2025-11-28 09:53:14 <tpbsd       > isene, the interesting issue about history is that most of it is innacurate
2025-11-28 09:53:50 <tpbsd       > take the internal combustion engine for instance. Did you know it was nearly outlawed ?
2025-11-28 09:54:40 <veltas      > I don't see a lot of cooperation between countries to avoid creating AI war horrors so probably the AI push won't stop until we're 2 seconds to midnight, if ever
2025-11-28 09:55:02 <tpbsd       > the reason is that until the internal combustion engine started powering cars (buggies) the kinds of injuries and deaths they caused had only ever been seen in wars
2025-11-28 09:55:12 <veltas      > What country would regulate AI when their adversaries can develop it freely and pose existential threats to them in future
2025-11-28 09:57:38 <tpbsd       > veltas, Im not worried about AI really because in general no country has the electricity and water to spare to power them as LLM's require prodigous amounts of both
2025-11-28 09:58:08 <tpbsd       > the chip industry is also competing for similar resources
2025-11-28 09:59:05 <tpbsd       > and my little 70GB home hosted AI is dumb and slow, it's not going to be a existential threat to anyone
2025-11-28 09:59:31 <tpbsd       > it needs $2000 in hardware to be dumb and slow
2025-11-28 10:02:04 <veltas      > Oh the current bubble has little fruit at all
2025-11-28 10:02:22 <veltas      > But one day there will be breakthroughs leading to general AI, in a week or 100 years
2025-11-28 10:02:47 <veltas      > And that scares me honestly
2025-11-28 10:02:55 <tpbsd       > yeah, if man can inagine the concept, it will exist one day
2025-11-28 10:03:27 <tpbsd       > whatever we imagine, eventually comes true
2025-11-28 10:04:03 <veltas      > They have constructed a mega cerebellum, capable of processing and categorising almost anything in general knowledge, but unable to really comprehend or think about anything
2025-11-28 10:04:14 <tpbsd       > true
2025-11-28 10:04:48 <veltas      > And they've done this with near limitless money and resources, all of human resources are being poured into it to the extent that our relatively power efficient server farms are now draining countries of electricity
2025-11-28 10:04:52 <veltas      > and still no dice
2025-11-28 10:04:57 <veltas      > So this clearly isn't going anywhere
2025-11-28 10:05:12 <veltas      > The breakthroughs won't be pure cash and farms of GPUs, it will be someone with a pen and paper
2025-11-28 10:05:15 <tpbsd       > only to a bubble that may pop soon
2025-11-28 10:05:49 <veltas      > I guarantee the person who 'contributes' the major breakthroughs will not receive any of the money for it, too
2025-11-28 10:05:58 <tpbsd       > yeah, also true
2025-11-28 10:06:04 <veltas      > There are patterns in history lol
2025-11-28 10:06:11 <tpbsd       > be some geek idiot
2025-11-28 10:06:17 <veltas      > genius
2025-11-28 10:06:24 <veltas      > But a geek
2025-11-28 10:06:40 <veltas      > Money isn't everything
2025-11-28 10:06:58 <veltas      > Less people think so every day, but it doesn't make it untrue
2025-11-28 10:07:26 <tpbsd       > as they say, when the guilotine blade seized just before the engineer was due to be beheaded, he looked up and said 'oh I see the problem, get me some axle grease and i'll soon fix that'
2025-11-28 10:07:38 <veltas      > lol that's good
2025-11-28 10:10:42 <isene       > Guys, this is enlightening as a discussion. I wrote a short story about the end of the world - an alternative explanation of the Fermi Paradox... basically that the reason we don't see aliens is scale of economics. It's easier and cheaper to go into a "matrix" than to travel to Alpha Centauri.
2025-11-28 10:14:04 <tpbsd       > isene, I watched a interesting YouTube the other day where an astronomer wondered if in fact our entire universe is inside a single black hole (he had lots of good reasons for his theory)
2025-11-28 10:15:30 <isene       > I asked that question of a professor back in 1980, and we had an interesting discussion about that
2025-11-28 10:15:48 <tpbsd       > personally I think that the greatest irony of mankind is that we think we matter, that we are 'intelligent', special etc. However we are related to everything else on this planet in some way, we are just the top of the food chain, thats all
2025-11-28 10:16:10 <isene       > I did the calculation back then, and it could work as a theory
2025-11-28 10:17:13 <tpbsd       > in reality we are at the mercy of the reality of the universe, and we have no say in anything really, other than our own lives and even then not a lot of say there either
2025-11-28 10:18:07 <tpbsd       > a huge meteor could wipe us all out in 10 minutes while we are discussing this stuff
2025-11-28 10:19:11 <tpbsd       > or nuclear war, or some new virus, or whatever
2025-11-28 10:19:35 <tpbsd       > or maybe we just all starve after wasting the limited resources here
2025-11-28 10:20:13 <tpbsd       > it's a roll of the dice, chaos reigns, and but for the grace of god go I
2025-11-28 10:20:21 <isene       > I have an alternative view which is pretty much 180 degrees opposite to that ;-)
2025-11-28 10:20:34 <tpbsd       > what is it ?
2025-11-28 10:20:52 <tpbsd       > Im 71, so all these theories interest me
2025-11-28 10:24:06 <isene       > I have researched the mathematics related to free will since around 40 years, and this summer I finally got the break-through, using Category Theory as a basis. I've written an academic paper on this that will be released in a few weeks and a more layman book that I'm looking for a publisher for
2025-11-28 10:25:16 <tpbsd       > I was thinking perhaps you may have a theory that we are all just simulations
2025-11-28 10:25:39 <isene       > No, it's more fundamental than that
2025-11-28 10:52:04 <ajhidd      > Some day people will understand that 'AI' is not worth all the water it wastes, but it may be too late
2025-11-28 11:03:11 <tpbsd       > ajhidd, I cant agree as I find my self hosted LLM very handy for certain tasks
2025-11-28 11:03:43 <tpbsd       > ajhidd, it's a tool like all my tools, it has pros and cons
2025-11-28 11:04:36 <tpbsd       > I think anyone who writes off AI 100% doesnt understand the toolk use anymore than anyone who promotes AI 100%
2025-11-28 11:04:40 <tpbsd       > it's just a tool
2025-11-28 11:05:20 <tpbsd       > I agree that in it's current form it does use a lot of water and that needs to be adressed
2025-11-28 11:05:28 <tpbsd       > and a lot of power
2025-11-28 11:07:16 <ajhidd      > And for what? To write a letter? Generate some sloppy code? They say that farming is bad for planet because cows produce CO2 that heats the planet. But what about these massive Ai data centers? Of course they are all good and green
2025-11-28 11:08:07 <ajhidd      > Yes, it's just a tool. But a tool that is not worth it. Like hammering a nail with a microscope
2025-11-28 11:12:00 <isene       > From what I have seen in discussions about AI, most arguments are emotional. People WANT it to succeed or WANT it NOT to succeed - mostly on emotional grounds. That worries me. Emotions can blind the actual issues at hand.
2025-11-28 11:18:16 <ajhidd      > The actual issue is AI's evergrowing enegry and water usage. It's not emotional - it's rational
2025-11-28 11:19:42 <isene       > That part is. The AI is "just crap, creating sloppy code", isn't.
2025-11-28 12:19:15 <tpbsd       > I agree with isene, Ive observed that the issue of AI is generally emotive and mainly financial, I guess like much in life
2025-11-28 12:19:57 <tpbsd       > us humans are always so emotive about everything
2025-11-28 12:20:43 <tpbsd       > and ajhidd is not wrong either, massive resources are being utterly wasted on AI
2025-11-28 12:20:58 <tpbsd       > but then nothing new there
2025-11-28 12:21:40 <tpbsd       > and the funny thing ? none of us here can affect what will happen anyway
2025-11-28 12:22:45 <isene       > This is a nice channel. Think I'm hanging around here
2025-11-28 12:24:21 <tpbsd       > and the chip fabs are using massive amounts of water also. I believe each chip used 1000 L of water to make
2025-11-28 12:24:45 <tpbsd       > farming uses massive amounts of water, far more than chips or AI
2025-11-28 12:25:38 <tpbsd       > isene, it is nice here, Forth people are generally pragmatic, smart, minimally emotive when debating
2025-11-28 12:26:10 <tpbsd       > isene, of course there are exceptions at times, like anywhere
2025-11-28 12:26:45 <isene       > Nice. If I ever come across as argumentative or condescending, then just tell me, ok?
2025-11-28 12:28:09 <tpbsd       > haha, who am I to judge ? Im just happy people chat with me
2025-11-28 12:30:31 <isene       > lol
2025-11-28 13:01:19 <veltas      > I've discussed on here before that you can measure the distance most of the way to Proxima Centauri to cm precision with 64-bit fixed point integers
2025-11-28 13:01:36 <veltas      > Or something like that, can't remember the exact figure but it's an easy calculation for anyone that wants to reproduce it
2025-11-28 13:02:15 <veltas      > Because in Forth we wax lyrical about fixed point
2025-11-28 13:03:53 <veltas      > Looks like you can measure the whole distance to Proxima Centauri in cm and beyond
2025-11-28 13:05:50 <veltas      > 3.83e16 m to proxima, 9.22e16 max 64-bit signed metres if denominating in cm, if I've calculated right, I guess Claude or a wet membrane can double check
2025-11-28 13:34:02 <isene       > Hehe. Cool.
2025-11-28 14:32:17 <KipIngram   > It's right at 4e18 centimeters to Proxima Centauri.  4e18/(2^64) is 0.2168.  So, yeah, well under a centimeter precision.  Even if you went with signed numbers you'd still be under half a centimeter.
2025-11-28 14:33:14 <KipIngram   > This is precisely why I see double precision integers as useful in a 32 bit system but not so much in a 64 bit system.
2025-11-28 14:33:32 <KipIngram   > They just feel unnecessary for practical problems to me.  Or have so far at least.
2025-11-28 14:34:48 <KipIngram   > You also see this in addressing.  You can fairly easily imagine using up your address space in a 32-bit computer, but a 64-bit address space is just... ridiculously gargantuan.
2025-11-28 14:35:13 <veltas      > Floating point is just better at automatically handling precision, but if you ignore precision you'll probably introduce error at some point anyway
2025-11-28 14:35:32 <veltas      > Although floating point also tends to come with tools to automatically detect this i.e. the rounding flags
2025-11-28 14:35:50 <veltas      > That said I've never been too interested in using floats in forth
2025-11-28 14:35:53 <KipIngram   > Right.  Floating point is more a convenience than anything else.
2025-11-28 14:36:22 <veltas      > I'd say more than a convenience, if you're going to have a floating exponent then you may as well use hardware floats, it will be more efficient
2025-11-28 14:36:34 <KipIngram   > And it's "what it is" - it's there and you get it even when you don't really need it.  With fixed point you control the details fully.
2025-11-28 14:36:53 <KipIngram   > Well, sure - I agree with that.  I just meant conceptually.
2025-11-28 14:37:21 <veltas      > Floats have been around since the very beginning of electronic computers, but there's plenty of older or embedded machines where floats aren't supported natively
2025-11-28 14:37:30 <veltas      > As an embedded engineer I don't like to overuse floats
2025-11-28 14:38:03 <KipIngram   > I saw a video recently about an alternate floating point format - I forget the name now but if  one of you say it I'll probably recognize it.  The presenter made a strong case that it was better than the standard floating point, but of course it would be awfully hard to change now because of what you just said - you can certainly use the new format if you want to, but then you lose that hardware
2025-11-28 14:38:05 <KipIngram   > performance advantage.
2025-11-28 14:38:17 <KipIngram   > Term for it started with "p," I think.
2025-11-28 14:38:31 <veltas      > I've not heard of this
2025-11-28 14:38:51 <KipIngram   > I'll see if I can find it and send it to you.  It was damn interesting.
2025-11-28 14:38:58 <veltas      > I like the standard floats, I think they're designed well
2025-11-28 14:39:03 <veltas      > Yeah I'd watch that
2025-11-28 14:39:04 <KipIngram   > May take me a bit - I'll  have to dig through my YouTube history.
2025-11-28 14:39:30 <KipIngram   > I'll wait a few minutes in case someone just says it, but I'll try to find it this morning.
2025-11-28 14:40:13 <veltas      > It's okay if it's too much effort
2025-11-28 14:40:17 <veltas      > I can live without knowing lol
2025-11-28 14:40:52 <tpbsd       > Ive just started a new project with a new mcu (new to me) that has a FPU and Ive only ever used fixed point until now, so I have some learning to do
2025-11-28 14:41:08 <veltas      > KipIngram: posits?
2025-11-28 14:41:20 <veltas      > tpbsd: Which MCU?
2025-11-28 14:41:28 <KipIngram   > Posits.
2025-11-28 14:41:30 <KipIngram   > Yes.
2025-11-28 14:41:41 <veltas      > Copilot helped me find it
2025-11-28 14:41:48 <veltas      > That's the sort of thing LLMs are good at
2025-11-28 14:42:00 <KipIngram   > I asked ChatGPT.  It coughed it right up, and all I needed was to see the word.
2025-11-28 14:42:18 <veltas      > Definitely a strong area for LLMs
2025-11-28 14:42:31 <tpbsd       > STM32G030C6
2025-11-28 14:42:32 <KipIngram   > The guy showed various types of calculations done both ways, across the entire range of values, and showed that in "most" cases posits offered superior accuracy.
2025-11-28 14:42:41 <KipIngram   > Not every single one - in a few corner cases posits lost.
2025-11-28 14:42:45 <KipIngram   > But only a few.
2025-11-28 14:42:56 <tpbsd       > 64KB flash 8 KB ram
2025-11-28 14:43:36 <KipIngram   > Anyway, I regarded the hour or so I spent on that video as wells pent.
2025-11-28 14:43:52 <tpbsd       > veltas, I've mainly been using the STM32F051 since 2014
2025-11-28 14:44:12 <KipIngram   > https://www.youtube.com/watch?v=aP0Y1uAA-2Y&t=3360s
2025-11-28 14:44:52 <veltas      > tpbsd: Are you sure it's got an FPU?
2025-11-28 14:44:54 <KipIngram   > You can install a posits package in julia if you want to play with them, but if your julia is new it fights you a little - I had to go in and adjust the package innards a little.
2025-11-28 14:45:05 <tpbsd       > veltas, the PDF says it does
2025-11-28 14:45:32 <KipIngram   > But once I coaxed it in it seemed to work correctly.
2025-11-28 14:45:54 <veltas      > Datasheet doesn't mention floating point, double, FPU; or am I missing something?
2025-11-28 14:46:02 <veltas      > It's M0+ which I think is just hardware integer multiply
2025-11-28 14:46:53 <tpbsd       > veltas, OOPS! brain fart, youre right!
2025-11-28 14:47:17 <KipIngram   > I think the best we can hope for is that someone MIGHT offer hardware support for conventional floats and posits at some point, but that would raise their cost, so... seems a bit like a long shot, given that it's not an already highly embraced thing.  It's a catch 22 - we can't really get the demand without the support, and it's the demand that will produce the support.
2025-11-28 14:47:19 <tpbsd       > veltas, I must have been looking at the PDF of another model\
2025-11-28 14:47:23 <veltas      > Looks like you don't need to learn floating point then lol
2025-11-28 14:47:48 <veltas      > Saved from it for a bit longer
2025-11-28 14:47:57 <tpbsd       > veltas, yeah, phew! I'm very happy with fixed point anyway, and use it all the time
2025-11-28 14:47:58 <KipIngram   > Basically I think they just make the internal breakdown of what's the mantissa and what's the exponent and so on a little more flexible.
2025-11-28 14:48:31 <veltas      > I work on a Cortex M4F core and I really think the floating point is a bit pointless (pun not intended)
2025-11-28 14:48:57 <tpbsd       > veltas, as Mo and M0+ have all I need for the things I do, I may never need to worry about FP
2025-11-28 14:49:22 <veltas      > Only scientific calculations are sensor readings that you could do with 16x16->32 multiply just fine
2025-11-28 14:49:48 <tpbsd       > veltas, I have some and have used them just to see how fast they are, the M4 is a beast!
2025-11-28 14:50:14 <tpbsd       > veltas, aha, tht makes sense
2025-11-28 14:50:22 <veltas      > Yeah got like 120MHz although we don't have it clocked that high
2025-11-28 14:51:16 <veltas      > The first thing I'd miss on a smaller M0 chip would be RAM
2025-11-28 14:51:20 <tpbsd       > even my old STM32F051 does 108 MHz when overclocked :) but it draws 30mA instead of 3ma when I do it
2025-11-28 14:52:07 <veltas      > Interesting
2025-11-28 14:52:20 <tpbsd       > veltas, as I only run Mecrisp-Stellaris and that fits in 20kb flash and uses about 1.5kb ram, I always feel like I have plenty of resources
2025-11-28 14:53:02 <tpbsd       > veltas, just for fun, I ran a test STM32F051 at 108Mhz for a year, and it was fine
2025-11-28 14:53:45 <tpbsd       > but of course the flash wont write at that speed, reads fine tho, and ram is fine
2025-11-28 14:54:25 <tpbsd       > but it was just a experiment to see if the mcu survived, and it didnt seem to bother it
2025-11-28 14:54:38 <veltas      > Is it one of those chips that hides the flash details or do you do manual NOR operations to erase/program?
2025-11-28 14:54:53 <tpbsd       > that included some pretty high ausie summer temperatures also
2025-11-28 14:55:32 <tpbsd       > I just use the ops provided by Mecrisp-Stellaris
2025-11-28 14:56:16 <veltas      > Would it be possible to set a lower bus clock with that overclocked core clock?
2025-11-28 14:56:18 <tpbsd       > I dont need to do manual operations as thats what Mecrisp-Stellaris does anyway
2025-11-28 14:56:21 <veltas      > I wonder if that would help
2025-11-28 14:56:55 <veltas      > Not that you need 108MHz but just an interesting problem
2025-11-28 14:57:36 <tpbsd       >  maybe, but I was only stress testing the chip, my projects would probably be fine at a 1MHz clock
2025-11-28 14:58:05 <tpbsd       > in my world everything is glacial, switches, sensors etc
2025-11-28 14:58:20 <tpbsd       > speed is the last of my needs
2025-11-28 14:58:32 <veltas      > Yeah I think that's not too unique to you
2025-11-28 14:58:43 <tpbsd       > lol, back in the day a 1 mhz 6800 was all I ever needed
2025-11-28 14:59:00 <veltas      > I think embedded in general is like that, we just get those high core clock speeds almost for free with modern CPU design
2025-11-28 14:59:25 <tpbsd       > true, the embedded world interfaces with slow things usually
2025-11-28 14:59:41 <veltas      > Which lets me understand them better as I am a slow thing as well
2025-11-28 14:59:58 <tpbsd       > veltas, yeah, I'm living in the golden age of embedded now
2025-11-28 15:00:39 <tpbsd       > I even get on chip debugging thrown in for free now! remember what ICE used to cost ?
2025-11-28 15:01:22 <tpbsd       > great huge adaptors with ribbon cables everywhere and a CPU that plugged into it
2025-11-28 15:01:46 <tpbsd       > now for $0.60 I get built in SWD/JTAG
2025-11-28 15:02:31 <veltas      > "remember what ICE used to cost" I don't know what that is, I'm pretty new to embedded lol
2025-11-28 15:02:39 <tpbsd       > oh!
2025-11-28 15:02:45 <veltas      > I do know lots of random bits of history but didn't live through it
2025-11-28 15:03:09 <veltas      > I'm in my 30s and have only really been working in embedded the last 3 years or so
2025-11-28 15:03:13 <tpbsd       > In Circuit Emulators: we had to use those in the 'old days' to do debugging
2025-11-28 15:03:21 <veltas      > Before then I was writing bare metal stuff but firmware for proper CPUs
2025-11-28 15:03:29 <veltas      > Bootstrap firmware etc
2025-11-28 15:03:46 <veltas      > Right I have seen those around actually
2025-11-28 15:03:47 <tpbsd       > oh, sorry, I didnt realise ...
2025-11-28 15:03:53 <veltas      > I can imagine those were expensive yeah
2025-11-28 15:04:19 <tpbsd       > well you entered embedded at the right time, those M4's are to die for
2025-11-28 15:04:34 <tpbsd       > so cheap, so packed with features
2025-11-28 15:04:52 <veltas      > As a hobby I wrote a Forth for ZX Spectrum so I do have some experience with older CPUs
2025-11-28 15:05:19 <tpbsd       > I have some H7xx's in 100 pin packages but havent used them yet, they have 90 or so on board peripherals!
2025-11-28 15:05:34 <veltas      > Crazy
2025-11-28 15:05:43 <tpbsd       > it amazing
2025-11-28 15:06:16 <tpbsd       > I sarted with the national Semi 'PACE' cpu in 1976, it was a 16 bit cpu
2025-11-28 15:06:40 <veltas      > Nice
2025-11-28 15:06:59 <tpbsd       > something like 48 instructions, paper tape for program development, teletype for terminal
2025-11-28 15:07:19 <tpbsd       > and printouts, not monitors
2025-11-28 15:08:17 <tpbsd       > the project was a iron ore flow guage that used radioactive isotopes to measure the shape of a falling curtain of iron ore
2025-11-28 15:08:41 <tpbsd       > and it calculated the flow rate in tons/time
2025-11-28 15:09:19 <tpbsd       > that was my introduction to embedded as a electronics technician, Im not a engineer
2025-11-28 17:14:14 <veltas      > What happens when you want the seahorse emoji in your LLM forth though