Hacker News Reader

Comment by bawolff | Hacker News Reader

Personally i'd rather we bridge the gap between using excel and writing python with something that is between the two, rather than relying on LLMs.

Call me a skeptic, but i remain very unconvinced that LLMs will be the enabling tool that lets non programmers program.

View on HN

> Personally i'd rather we bridge the gap between using excel and writing python with something that is between the two, rather than relying on LLMs.

Maybe Lua[1] and Excel. Python's significant whitespace is unsuitable for doing the one-liners that spreadsheet experts do.

Writing code for the spreadsheet should not require putting that code in a separate file just so the significant whitespace can be maintained.

Whatever you write in a single cell should be the programming language. If you later want to move it to it's own file, module, package, CLI, etc, you should be able to.

[1] Visual Basic turned out to be very usable with spreadsheet experts.

Excel's cramped formula bar that defaults to one line is, to borrow a turn of phrase from Raskin, not at all something that is necessary—it is merely customary.

There's no (good) reason that when you start to type a formula or you click on a cell that has a formula already in it that Excel doesn't make an ephemeral sidebar overlay appear with an 80-column text editor in it. There's no good reason that it doesn't run a gofmt-style code prettifier on long, sprawling formulae. There's no good reason that clones like Google Docs don't do this, either. The best time would have been when we transitioned to widescreens.

Spreadsheet programming is bad because neither Microsoft nor its competitors actually care about making it good.

I agree that something between Python and Excel is a good idea. But still, LLMs are good at using in context learning to convert natural language to a given pattern language or format, and where people don't know what the syntax is. Making visual interfaces that are so easy to use that they don't need to use such magic would be nice. But it seems hard to do that.

Microsoft has gone and done it:

https://support.microsoft.com/en-us/office/get-started-with-...

There is visidata. It's a TUI and still has rough edges, but it could develop into what you're envisioning.

How would this new intermediate be different from SQL?

Well, if i knew what the solution was i would be making it instead of waxing poetic about it on hn.

While sql (or say microsoft access) could be seen as an intermediary, i don't really think it is. It is very specialized to specific applications, and arguably rdbms are more complex than basic python.

Reply on news.ycombinator.com