I have tried many many tools over the years. Each time limitations, platform support or proprietary data formats frustrated me. In recent years I started to experiment just using Markdown files, but then cross references and having to organise document and folder structure ahead of time proved to be yet another failing.
Started to look at a few modern takes on the problem including Roam Research. A few months ago I found https://logseq.com/ Still in beta but progressing quickly. So far I think I have finally found my happy place.
I have had a few cards go but I never isolated the cause. Certainly unceremoniously cutting power occurred a few times.
I got tired of the inconvenience of having to rebuild the Pi installation and flashing SD cards, so now I boot my Pi4 directly from a USB attached SSD, no card on-board at all. I haven't looked back. Others seem to also be getting good result like: https://www.jeffgeerling.com/blog/2020/im-booting-my-raspber...
There is also a difference between buggy for the first time, which tends to cause a user to dismiss the app, versus sometimes buggy after the app is well established as useful for the user. At that point, there is a different dynamic at play for competitors, they need to be perceived a significant percentage better than the incumbent to overcome and justify the effort of the user changing. I have no sources to back that up sorry, just something observed.
The pieces of Snips that were OSS before Sonos bought and ditched the community are being leveraged along with new work in Project Alice https://github.com/project-Alice-assistant It continues to strive to be modular and offline. By design, the choice of online/offline elements including Google ASR and Amazon TTS along with corresponding quality and privacy tradeoffs is your choice. Come give a hand.
To be fair there are others that have been pushing the needle in this space for considerable time. The standout for me is https://snips.ai Offline, multiple platforms, multiple languages, many parts open source and more oss parts in the pipeline. While certainly not currently aimed at dictation, but instead assistant building and automation. In this space on device speed, privacy & offline are critical. In the case of Snips "piss poor" falls short from the reality of what I have experienced YMMV.
I made a fulltime living between 1993 - 2005 using VizualWorks which is the original descendant Smalltalk from ParcPlace which came out of Xerox PARC. Systems are still in production across industries. It was a particular favourite with banks and Wall St for decades and some like JP Morgan still use it http://www.cincomsmalltalk.com/main/successes/financial-serv...
I saw JPMorgan on the Pharo Consortium list. Any idea how they use it? I get the feeling large banks use almost EVERYTHING as they even have a lot of APL language use in house with A+ & kdb+.
Actually I was surprised to see JP Morgan there. I don't watch it closely but they seem a recent addition. Their "Kapital" system is based on Cincom Smalltalk. I wouldn't speculate they are porting it. Facilitating a viable open source Smalltalk alternative may just for negotiating power with their incumbent Smalltalk.
http://www.cincom.com/pdf/CS040819-1.pdf
I met the International Product Manager for Cincom in a Smalltalk meetup once. She was complaining about the aggressive Pharo folks "badmouthing" Cincom and taking their business. One of the chip fabs previously running VW Smalltalk got taken by Pharo. I heard somewhere else that JPM were unhappy at the steep consultancy fees Cincom was charging, and the lack of VW Smalltalkers out there, but still wanted to keep things in Smalltalk. So it seems these rumours may be true if JPM has joined the Pharo Consortium.
I was wondering the same thing as I figured they already used Cincom. I know Cincom is very expensive, but surely it's negligible for JPMorgan? Maybe they feel Pharo has more momentum and have started doing some projects there to steer it where they want it to go?
How often are you updating your results ? Can I download the recommendations dataset for offline queries ?