Interested. I got far emough on the site to have a couple of questions. (Maybe the answers are on the site?)
1) Does device record ALL MIDI sent to it? (Including e.g. SYSEX and non-registered parameters?
2) I edit MIDI on a Desktop (NOT a phone).
a) Are sessions recorded into separate files? Named and/or dated?
b) Can I inspect & load sessions from the card?
How did you navigate the world of compliance testing and certification, with all the different countries involved (FCC, UL, etc). Are there any markets you've decided not to sell to as they're not worth the paperwork?
I'm an EE with main expertise in IC design but I design a lot of the chip packages, PCBs, mechanical stuff like custom made sockets etc and write a ton of SW, be it tools for myself for development and testing, be it the firmware of the chip I'm designing or another chip I use on the board. For me it never felt like SW was the hard bit, because you can write and test it pretty immediately and pretty much for free. HW felt harder because of the tooling costs, lack of documentation and examples, things that aren't obvious at the beginning (EMI/EMC) etc etc but these are very minor compared to the actual issue: long and expensive development cycles. Now, that doesn't apply if your HW is extremely simple and just a glorified uC and rhe development time is mostly the code running on it. This is why this is such a poor take. One can argue for anything at extremes. SW can be very very complex and much harder than HW, and vice versa at different extremes.
Wow, congrats, I remember reading your HN post. I hope you will continue writing blog posts on your journey, I expect quite a lot of software engineers would be very interested going though a similar journey now that LLMs removed most of the fun of programming. It’s neat to work on something physical for a change
> quite a lot of software engineers would be very interested going though a similar journey now that LLMs removed most of the fun of programming.
I suspect we will see a reversal on this. LLMs are flat out worse than humans at programming (yes, even your latest favorite frontier model). I am faster (once you factor in the time fixing its inevitable mistakes) and better at programming than Claude is, and I'm not a particularly exceptional programmer. People say "they aren't going away", but I disagree with that - already we've seen people disenchanted with the lack of results compared to what they pay, and that is only going to grow once the AI companies raise prices in an attempt to turn a profit rather than lock people in. Eventually the industry will wake up and stop drinking the kool-aid, and what matters will once again be the results you deliver, not the tools by which you deliver them. But it's hard to say when that will happen, and it's going to be rough in the meantime.
>>> The hardest part of building Jamcorder was still, by far, the software -- roughly 200K lines of code spread across the firmware, app, and manufacturing tooling. It took over 3 years and many long nights in a pre-LLM world.
>>> When compared to that, the hardware was undeniably smooth sailing.
>>> For the record, I don’t think I’m special. It’s just that hardware’s reputation for being difficult is, IMO, overstated.
For me, when I complain about hardware, its actually usually complaining about drivers, compatibility, underdocumented interfaces, firmware etc. If one were to put firmware and manufacturing tooling in a separate bucket, that is indeed the problem!
Ok are you riding with two huge boxes containing all your work on an electric unicycle and I guess the hi-vis top means you're headed into traffic and the box-cords seem to be balancing at the halfway point of the boxes and this is too much anxiety for me. But hey, glad you optimized for minimalism on the PCB ♡
Very happy JamCorder customer here. I've started playing in improv shows since I bought it too, and it's nice to have those on record!
It's rare that I come across a product I have literally zero complaints about, or any feedback to offer. Even my concern that it was tied to an app that might disappear one day is of course covered by the fact that it's all just MIDI files on a card.
Thank you so much for making this, it's basically a perfect product for me and I love it. No notes.
>Even my concern that it was tied to an app that might disappear one day is of course covered by the fact that it's all just MIDI files on a card.
I was sorta surprised when the blog mentioned it was like 200k lines of code for an app. I wouldn't think you'd even need an app for something that puts midi files on an sd card.
With 2500 units sold, no one has moved to counterfeit yet(1). A device this simple will be cloned by AI in an afternoon. There is no imaginable anticounterfeit strategy.
(1)2500*$175=$437,500. Estimated profit is 33-66%, so $150-300k. Nice for a soloprenuer, but limited market, so small target for piracy.
Hardware isn’t hard if the hardware is easy? There’s very few components, one of which is a pre made module, and it took 3 years to make essentially a text logger.
Congrats on an impressive achievement and milestone. This is a great outcome, great work, and an indication of how far the hardware industry/tooling has come.
This feels though a little to me like building a simple SaaS service, and deploying it on a managed cloud service, and concluding all software is really easy!
This project is near to (if not at) the easy end of the hardware product difficulty/complexity spectrum. Some great choices were made to keep it simple and therefor easy.
An alternate take-away: "If you keep things simple, hardware doesn't have to be as hard"
Congratulations! I bought one after reading your post on here announcing it. It's fun to mess around on my keyboard and be able to go back and listen to what I played. Great for learning
102 comments
[ 0.85 ms ] story [ 22.0 ms ] thread1) Does device record ALL MIDI sent to it? (Including e.g. SYSEX and non-registered parameters?
2) I edit MIDI on a Desktop (NOT a phone). a) Are sessions recorded into separate files? Named and/or dated? b) Can I inspect & load sessions from the card?
I'm an EE with main expertise in IC design but I design a lot of the chip packages, PCBs, mechanical stuff like custom made sockets etc and write a ton of SW, be it tools for myself for development and testing, be it the firmware of the chip I'm designing or another chip I use on the board. For me it never felt like SW was the hard bit, because you can write and test it pretty immediately and pretty much for free. HW felt harder because of the tooling costs, lack of documentation and examples, things that aren't obvious at the beginning (EMI/EMC) etc etc but these are very minor compared to the actual issue: long and expensive development cycles. Now, that doesn't apply if your HW is extremely simple and just a glorified uC and rhe development time is mostly the code running on it. This is why this is such a poor take. One can argue for anything at extremes. SW can be very very complex and much harder than HW, and vice versa at different extremes.
What understanding about hardware did you start with, and what tools did you use to learn what you didn’t already know?
https://www.izcorp.com/wp-content/uploads/pm_Pocket-Record.j...
I suspect we will see a reversal on this. LLMs are flat out worse than humans at programming (yes, even your latest favorite frontier model). I am faster (once you factor in the time fixing its inevitable mistakes) and better at programming than Claude is, and I'm not a particularly exceptional programmer. People say "they aren't going away", but I disagree with that - already we've seen people disenchanted with the lack of results compared to what they pay, and that is only going to grow once the AI companies raise prices in an attempt to turn a profit rather than lock people in. Eventually the industry will wake up and stop drinking the kool-aid, and what matters will once again be the results you deliver, not the tools by which you deliver them. But it's hard to say when that will happen, and it's going to be rough in the meantime.
Sounds like a story. What parts of it can you share?
>>> The hardest part of building Jamcorder was still, by far, the software -- roughly 200K lines of code spread across the firmware, app, and manufacturing tooling. It took over 3 years and many long nights in a pre-LLM world.
>>> When compared to that, the hardware was undeniably smooth sailing.
>>> For the record, I don’t think I’m special. It’s just that hardware’s reputation for being difficult is, IMO, overstated.
For me, when I complain about hardware, its actually usually complaining about drivers, compatibility, underdocumented interfaces, firmware etc. If one were to put firmware and manufacturing tooling in a separate bucket, that is indeed the problem!
It's rare that I come across a product I have literally zero complaints about, or any feedback to offer. Even my concern that it was tied to an app that might disappear one day is of course covered by the fact that it's all just MIDI files on a card.
Thank you so much for making this, it's basically a perfect product for me and I love it. No notes.
I was sorta surprised when the blog mentioned it was like 200k lines of code for an app. I wouldn't think you'd even need an app for something that puts midi files on an sd card.
The app is optional, and its for browsing your library. Think of it like the Photos app but for MIDI.
Also, pretty cool that you set up your website real fast. Would you mind adding your Twitter link onto the site?
(1)2500*$175=$437,500. Estimated profit is 33-66%, so $150-300k. Nice for a soloprenuer, but limited market, so small target for piracy.
This feels though a little to me like building a simple SaaS service, and deploying it on a managed cloud service, and concluding all software is really easy!
This project is near to (if not at) the easy end of the hardware product difficulty/complexity spectrum. Some great choices were made to keep it simple and therefor easy.
An alternate take-away: "If you keep things simple, hardware doesn't have to be as hard"