I've heard about Improv for forever, but NeXT emulators didn't exist and I couldn't afford even a very used NeXTcube at the time. It was really wonderful seeing Improv's operation and benefits so clearly listed out here, and made me genuinely wonder how hard it'd be to make a version of this in 2026
Amazing! Like the poster, playing with NeXT is something I'd love to do. (There are browser-based emulators at https://infinitemac.org/?filter=next.)
But this story combines bot NeXT and "alternatives to spreadsheets", which is another of my favorite topics!
Spreadsheets are one of those "worse is better" technologies or whatever you want to call them. Many of us see they have unavoidable huge defects, but a lot of people will use spreadsheets and will not switch to anything else.
I'm a relational database fan, and I've been waiting forever for an easy-to-use relational database to replace spreadsheets. I've been waiting so long that I should know better it might never happen. I thought Notion and similar products would move the needle, but Coda (2019) and Notion (2016) have been out for a while and it doesn't seem to be happening.
If a squint a bit, Improv looks a bit like a relational database but with a bit of hierarchical orientation.
Some day we will manage to do something better than Visicalc (1979). Funny how we can build some amazing software, but spreadsheets remain- and stuff such as writing CRUD apps are still in the stone ages, IMHO.
I've been waiting so long that I should know better it might never happen. I thought Notion and similar products would move the needle, but Coda (2019) and Notion (2016) have been out for a while and it doesn't seem to be happening.
If you squint, Notion resembles Semantic MediaWiki more than anything else, which is more of a graph database.
Some day we will manage to do something better than Visicalc (1979). Funny how we can build some amazing software, but spreadsheets remain- and stuff such as writing CRUD apps are still in the stone ages, IMHO.
I think it's that fundamentally spreadsheets give you immersive feedback, and to do that, they have to be technically impressive. They're a 2D dataflow livecoding environment, after all.
I like how Notion is "text documents with embedded structured data and logic", something with IIRC Microsoft Office had ages ago, but I never saw it very well used. I think notebooks, word processors, wikis should converge to something. I'd kinda add Git repositories and code review too.
And yes, Notion treats documents as entries in the database too, IIRC, which is pretty cool and makes it lean more towards that side of the equation.
But I think Notion tables (and Coda's equivalent) concept are more relational.
My own take on this was writing a Pandoc filter that dumped tables in a Markdown document into an SQLite database, and code blocks marked as SQL would be executed on that database and replaced by the resulting table. It was actually somewhat useful and nice.
I wanted to do a second version with Djot and Datalog but I never got very far with it.
In my head, it's you have a file tree like the ones that all those lightweight markup documentation tools use (Sphinx, mdbook, whatever). But you also keep tabular data in a nice text format. And your text documents can embed queries on that data and generate maybe not only tables, but perhaps charts or even other documents.
You can choose to use this from a web UI, but alternatively you can use VCS tools on it that enable code review and similar stuff (perhaps publishing to a GH Pages-like service, etc.).
But also maybe you have an interactive notebook interface with nice table editing and where you can iterate on the queries that generate charts and tables.
Not sure how to make that both multiplayer friendly (e.g. realtime CRDT-style live cursor collaboration) and also "code-review" async review friendly, though.
edit: also I think Improv (and a few others) have the "live" feedback too, but... flat grid seems to be what works best for people?
I think the problem is a lot of what programmers think would improve spreadsheets would involve basically up-front data decisions (static types, designing the grid ahead of time), taking away that immediate feedback and experimentation, at which point, you have a database with nice report UIs and not a spreadsheet. Which, a graphical database you could sell to the average person is also very valid - no one's been able to crack the Access nut in years.
I'm not so sure about that- Notion seems to handle it relatively gracefully, right? You can add, alter, and drop columns reasonably. Also "refactoring" spreadsheets is not a walk in the park either.
But I think what you say goes in the general direction. It's just easier to get started with spreadsheets. But maybe it's just that intuitive equals familiar? Also I think spreadsheets don't scale up at all.
So that creates a bad situation: switching to anything that is not a spreadsheet has a massive initial cost... and you don't need it in a ton of cases! Plus, there don't seem to be so many good alternatives to switch to.
I will say that I've seen (and been part of) projects of programmers to replace user spreadsheets with CRUD web apps (so not even the ideal formula case of spreadsheets, but an actual database) thing and they inevitably fail, even when asking the users what they want. The users made their own tooling that works.
The best I've heard of to try to square the circle is Bank Python with explicit concessions to make it easy as possible for users, lest shadow IT occurs.
Heh, if you want an opposite data point, I worked for a records company. They used spreadsheets for album metadata. This worked fine when uploading their music to services that accepted spreadsheets as metadata, but they hired me when Spotify demanded XML metadata.
I could have solved the issue with just a spreadsheet to Spotify XML converter, but I gave them a CRUD app. In my opinion, it was the right answer because they had other preexisting problems that in my opinion were due to using spreadsheets and unsolvable with spreadsheets.
(AFAIK, they were happy with the CRUD app for a few years. I don't think they're still using it, I expect someone finally developed an off-the-shelf product any records company could use.)
But I think you're right anyway. The problem is that most people can get much farther with spreadsheets than with anything else. And managing a software project (i.e. paying developers and getting the product you want) has always been broken. In my case I think they got lucky, but I think in general even with very favorable conditions for the product you need, it's a long shot that you'll get your CRUD app under reasonable budget and time. The other option is that your business is common enough that someone develops a nice off-the-shelf product that works for you.
I worked for a fintech that when I joined was ~4 years old. At that point they were transitioning from what basically amounted to several ginormous google sheets that were being transformed by dagster into other google sheets. It was sheets all the way down.
We migrated it into a CRUD and while in the beginning it was a struggle, it quickly became much better than that.
But that sustained a multi-million dollar business for 4 years.
My own take on this was writing a Pandoc filter that dumped tables in a Markdown document into an SQLite database, and code blocks marked as SQL would be executed on that database and replaced by the resulting table. It was actually somewhat useful and nice.
Thanks, I didn't know of that one. I never said it was an original idea; Logseq and Obsidian also have similar things. The main point of my implementation is that I could get something useful in 70 lines of Python :)
Oh I wasn't critiquing! More like, "if you still like this idea, this actively maintained project is really close". I stopped maintaining my own notes system a few months ago for similar reasons
Oh, that's cool- maybe I was too blunt. It's a topic I'm interested in and it's nice to look at other attempts to solve it. But it feels like spreadsheets are enough for so many purposes there isn't really a lot of chances for alternatives to thrive.
[OP] gecko | 11 hours ago
I've heard about Improv for forever, but NeXT emulators didn't exist and I couldn't afford even a very used NeXTcube at the time. It was really wonderful seeing Improv's operation and benefits so clearly listed out here, and made me genuinely wonder how hard it'd be to make a version of this in 2026
koala | 10 hours ago
Amazing! Like the poster, playing with NeXT is something I'd love to do. (There are browser-based emulators at https://infinitemac.org/?filter=next.)
But this story combines bot NeXT and "alternatives to spreadsheets", which is another of my favorite topics!
Spreadsheets are one of those "worse is better" technologies or whatever you want to call them. Many of us see they have unavoidable huge defects, but a lot of people will use spreadsheets and will not switch to anything else.
I'm a relational database fan, and I've been waiting forever for an easy-to-use relational database to replace spreadsheets. I've been waiting so long that I should know better it might never happen. I thought Notion and similar products would move the needle, but Coda (2019) and Notion (2016) have been out for a while and it doesn't seem to be happening.
If a squint a bit, Improv looks a bit like a relational database but with a bit of hierarchical orientation.
Some day we will manage to do something better than Visicalc (1979). Funny how we can build some amazing software, but spreadsheets remain- and stuff such as writing CRUD apps are still in the stone ages, IMHO.
calvin | 7 hours ago
If you squint, Notion resembles Semantic MediaWiki more than anything else, which is more of a graph database.
I think it's that fundamentally spreadsheets give you immersive feedback, and to do that, they have to be technically impressive. They're a 2D dataflow livecoding environment, after all.
koala | 7 hours ago
Oh well, that's another thing.
I like how Notion is "text documents with embedded structured data and logic", something with IIRC Microsoft Office had ages ago, but I never saw it very well used. I think notebooks, word processors, wikis should converge to something. I'd kinda add Git repositories and code review too.
And yes, Notion treats documents as entries in the database too, IIRC, which is pretty cool and makes it lean more towards that side of the equation.
But I think Notion tables (and Coda's equivalent) concept are more relational.
My own take on this was writing a Pandoc filter that dumped tables in a Markdown document into an SQLite database, and code blocks marked as SQL would be executed on that database and replaced by the resulting table. It was actually somewhat useful and nice.
I wanted to do a second version with Djot and Datalog but I never got very far with it.
In my head, it's you have a file tree like the ones that all those lightweight markup documentation tools use (Sphinx, mdbook, whatever). But you also keep tabular data in a nice text format. And your text documents can embed queries on that data and generate maybe not only tables, but perhaps charts or even other documents.
You can choose to use this from a web UI, but alternatively you can use VCS tools on it that enable code review and similar stuff (perhaps publishing to a GH Pages-like service, etc.).
But also maybe you have an interactive notebook interface with nice table editing and where you can iterate on the queries that generate charts and tables.
Not sure how to make that both multiplayer friendly (e.g. realtime CRDT-style live cursor collaboration) and also "code-review" async review friendly, though.
edit: also I think Improv (and a few others) have the "live" feedback too, but... flat grid seems to be what works best for people?
calvin | 7 hours ago
I think the problem is a lot of what programmers think would improve spreadsheets would involve basically up-front data decisions (static types, designing the grid ahead of time), taking away that immediate feedback and experimentation, at which point, you have a database with nice report UIs and not a spreadsheet. Which, a graphical database you could sell to the average person is also very valid - no one's been able to crack the Access nut in years.
koala | 7 hours ago
I'm not so sure about that- Notion seems to handle it relatively gracefully, right? You can add, alter, and drop columns reasonably. Also "refactoring" spreadsheets is not a walk in the park either.
But I think what you say goes in the general direction. It's just easier to get started with spreadsheets. But maybe it's just that intuitive equals familiar? Also I think spreadsheets don't scale up at all.
So that creates a bad situation: switching to anything that is not a spreadsheet has a massive initial cost... and you don't need it in a ton of cases! Plus, there don't seem to be so many good alternatives to switch to.
calvin | 5 hours ago
I will say that I've seen (and been part of) projects of programmers to replace user spreadsheets with CRUD web apps (so not even the ideal formula case of spreadsheets, but an actual database) thing and they inevitably fail, even when asking the users what they want. The users made their own tooling that works.
The best I've heard of to try to square the circle is Bank Python with explicit concessions to make it easy as possible for users, lest shadow IT occurs.
koala | 5 hours ago
Heh, if you want an opposite data point, I worked for a records company. They used spreadsheets for album metadata. This worked fine when uploading their music to services that accepted spreadsheets as metadata, but they hired me when Spotify demanded XML metadata.
I could have solved the issue with just a spreadsheet to Spotify XML converter, but I gave them a CRUD app. In my opinion, it was the right answer because they had other preexisting problems that in my opinion were due to using spreadsheets and unsolvable with spreadsheets.
(AFAIK, they were happy with the CRUD app for a few years. I don't think they're still using it, I expect someone finally developed an off-the-shelf product any records company could use.)
But I think you're right anyway. The problem is that most people can get much farther with spreadsheets than with anything else. And managing a software project (i.e. paying developers and getting the product you want) has always been broken. In my case I think they got lucky, but I think in general even with very favorable conditions for the product you need, it's a long shot that you'll get your CRUD app under reasonable budget and time. The other option is that your business is common enough that someone develops a nice off-the-shelf product that works for you.
marcecoll | 2 hours ago
I worked for a fintech that when I joined was ~4 years old. At that point they were transitioning from what basically amounted to several ginormous google sheets that were being transformed by dagster into other google sheets. It was sheets all the way down.
We migrated it into a CRUD and while in the beginning it was a struggle, it quickly became much better than that.
But that sustained a multi-million dollar business for 4 years.
[OP] gecko | 5 hours ago
You're floating around SilverBullet, just FYI
koala | 5 hours ago
Thanks, I didn't know of that one. I never said it was an original idea; Logseq and Obsidian also have similar things. The main point of my implementation is that I could get something useful in 70 lines of Python :)
[OP] gecko | 5 hours ago
Oh I wasn't critiquing! More like, "if you still like this idea, this actively maintained project is really close". I stopped maintaining my own notes system a few months ago for similar reasons
koala | 4 hours ago
Oh, that's cool- maybe I was too blunt. It's a topic I'm interested in and it's nice to look at other attempts to solve it. But it feels like spreadsheets are enough for so many purposes there isn't really a lot of chances for alternatives to thrive.
ploum | 8 hours ago
The URL is blocked by my ublock as being part of Phishing URL blocklist.
The rule is "||stonetools.ghost.io^"
Is this a false positive ?
[OP] gecko | 8 hours ago
False positive; see https://mastodon.sdf.org/@cosi/117427424421317363
ploum | 8 hours ago
Thanks! I wonder how you ends up being blocked that badly. (sometimes, it’s nothing more than a malware domain being too close to yours I guess)