I've already had a few people email me links to their model with questions about implementation. Since posting on HN there have been more than 2,000 models created. If someone finds a serious bug in a formula or better yet, a new feature is added then they all get the benefit. There's not very clear path(s) to upgrading an Excel workbook.
And my straw man argument: I could make a case for building a lot of stuff in Excel that currently exist as successful hosted applications. Lowering the barrier to entry is one way to differentiate.
And if you want to understand what you see, and tweak it to your business, you must have a very deep undestanding of all the mechanics behind the cells, which is very hard/painful/time consuming to obtain when you use a complex spreadsheet from someone else.
Still, I really like the concept of funded.io. Combining planning and tracking into one tool (automated for tracking) would be awesome!
Perhaps it's because I've spent many, many hours using Excel (although less so these days), but those problems you state don't seem like problems:
1) Fat-fingering is a problem only if you mix the task of coding (setting up the structure and formulae in Excel) with the task of using (changing the parameters). You don't code your web site at the same time as using it (except for manual testing), and it can be the same with Excel. Highlight all input cells with a yellow background and blue font, unset their 'Protected' property, and then lock the workbook. You/anyone can then play to your heart's content without worrying about messing up formulae.
2) When you are creating new versions of a spreadsheet, it's either because you are changing the structure (e.g. changing formulae, adding columns) or because you are changing the input data. If the latter then, yes, you can have several different versions. I've managed this in the past (when fundraising and doing regular reviews with investors) by having an 'Assumptions' tab which contains _all_ inputs. Each scenario (set of assumptions) has a name/number and a column of its own. The calculations and output dynamically reference just one scenario at a time depending on the scenario chosen from a drop-down. Not only that, but Excel's 'Data tables' feature (which explicitly supports only one-variable or two-variable scenarios) works with this, even if you have 10s of variables: just set the scenario number as your variable.
3) Using a 'template they found' is clearly a bad idea. They wouldn't just copy-paste some 'code they found' without studying it and the tests. In fact, for something simple, they'd probably just re-create their own after studying someone else's code for something similar. It's the same with Excel files.
4) Can't argue with this one :)
5) This last point intrigues me: until now I thought you were focused on business plans, and that you wouldn't overwrite the historical data with actuals. However, you mention operational data, suggesting you are talking about combining planning and tracking into one tool. Is this correct?