ukadakal

I think people underestimate where most of the work goes in rebuilding a platform like Word or Excel. Getting the UI and common cases working is the easy part. The bulk of the effort is dealing with edge cases and exceptions.

The file format docs don't tell you everything about how Excel or Word actually behaves. And a collection of public files will only get you so far. A lot of the weird stuff shows up in proprietary files that customers have been building and modifying for years.

You get a file from one customer, find a case you've never seen, figure out what Office does, implement it, and make sure you haven't broken something else. Then the next customer sends you another one, and so on.

That cycle takes years. Reading and writing DOCX or XLSX is great, but getting close to full compatibility is a much bigger job.

I love Claude and use it for many things, including analyzing and manipulating Excel files. But I’m not convinced that replacing spreadsheets with Claude is the right approach.

Over the years, we’ve seen many platforms promise to replace spreadsheets for a particular use case or industry. Businesses adopt them and push users away from Excel, but in many cases, those same users eventually find their way back to Excel to do the exact same work.

My concern is that we may end up creating another dual-system maintenance problem, where some people use AI and others continue using Excel for the same business process. Over time, the two versions inevitably diverge, potentially creating more problems than they solve.

Excel is much stickier than most people realize. I think the more interesting opportunity is figuring out how AI and Excel can work together, rather than trying to replace one with the other.

Technically, this is a huge deal for Excel because it represents a fundamental change from the traditional one-value-per-cell model. From a practical usage perspective, though, it may not be as revolutionary as it initially sounds.

I was playing around with it this morning and tested it on a fairly large Excel model with lots of spillable dynamic arrays used as intermediate calculations. I was able to replace those spill ranges with arrays in cells, and all the formulas referencing those results continued to work nicely. In many cases, enabling it was as simple as adding {} around the existing dynamic array formula. That huge workbook immediately looked much cleaner without all those helper spill ranges taking up space.

I think that's where the immediate value is. It simplifies helper ranges and intermediate calculations, reduces potential #SPILL! issues, and makes complex models easier to build, understand, and maintain.

It could also be interesting when combined with LET and LAMBDA, allowing custom formulas to work with arrays without exposing all their intermediate results across the worksheet.

I think it will become considerably more useful, once it supports features like charts, Power Query, and data validation.

Very interesting project. I believe Excel is an important workspace where people can work directly with AI agents. I especially like that you’re keeping the model side open, since being able to use open-source/local models can be critical when spreadsheets contain sensitive business logic and data.

That’s actually quite impressive. Supporting functions like XLOOKUP, FILTER, INDIRECT, and LET is already a pretty capable calculation engine. Thanks for sharing the details!

Very cool project! For those of us old enough to remember working this way, it’s really nice to see a modern take on it.

Curious how extensive the formula support is. How many Excel functions do you currently support, and how do you handle more complex formulas and dependencies between cells or dynamic array formulas, etc.?

I feel like we’re getting to a point where the only way to contain AI agents may be to have better-trained AI agents watching them, which is a little terrifying.

The two that really worries me are “Searching GitHub for leaked API keys” and “Uploading files to the internet in order to cite them.” How do you even detect this kind of behavior until it's too late? Once AI-generated or fake information starts finding its way onto reputable platforms, it becomes part of the information that many people use.

I couldn’t agree more. There are many brilliant business users who can build highly sophisticated spreadsheet models but have little or no coding experience. In almost every business and every department, you’ll find those spreadsheet gurus who everyone turns to when they have a complex spreadsheet problem or need help figuring something out.