← Back to the main page

Simon, building Betty.

Notes on building Betty, writing books, fixing strange edge cases, and figuring out where this whole thing should go next.

About that last 20%.

In May I wrote a post here called "The last 20% took 80% of the time." I was rather pleased with the insight. I had not, it turns out, finished the last 20%. I had finished about half of it.

Read more

In May I wrote a post here called "The last 20% took 80% of the time." I meant it as a dispatch from the finish line. Four months on, I can report that it was a dispatch from somewhere near the middle.

This is a very authorly thing to have done. Every writer I know has announced they were nearly finished with a draft, and every one of them has been wrong, and not one of us has learned anything from it. I have now managed it about software, which is at least a change of genre.

So, in the spirit of the thing, here is what "nearly done" turned out to contain.

I built a benchmark. The idea is simple enough: take clean manuscripts, plant errors at known places in four languages, and count what Betty actually catches. The first thing it told me was that I had been guessing about how good she was. The second was that I had run an entire round of it with the grammar checker switched off without noticing, which meant four days of results were worth nothing and one of my conclusions was exactly backwards.

Then I spent a week arguing with myself about commas. Betty scores badly on them, and I had a very comfortable theory about why: commas are a matter of taste, so failing to reproduce one author's choices isn't really an error. I liked that theory a lot. It was doing me a favour. So I went and measured it properly instead, and it turned out to be half right — and the half that survived pointed at a genuine fault I hadn't known about. That is roughly the story of this whole summer.

There is also now a way to run Betty without a powerful computer, on servers in France that keep no copy of your writing. That was not in the plan in May either. Most things weren't.

None of this is the interesting part of building a writing tool. All of it was necessary, and none of it was visible from where I was standing in May — which, I suspect, is the actual lesson, and not the tidy one about percentages I thought I had learned the first time.

So I am not going to tell you I'm nearly done. Betty works, she is out, the numbers are published including the unflattering ones, and I will go on finding the next 20% for as long as it keeps turning up.

If you want to see what she catches and what she doesn't, it's all on the performance page. The comma row is still the worst one there. I've stopped pretending that's fine.

— Simon

The last 20% took 80% of the time.

Building Bethaniel has felt a lot like writing a book: you think you're nearly done, and then the final stretch expands until it eats your calendar.

Read more

Building Bethaniel has felt a lot like writing a book: the last 20% took 80% of the time.

I thought I was in the final stretch ages ago. Then I spent what feels like forever fixing how the system recognizes scene breaks and paragraph breaks, and tuning performance so Betty feels less like a clever prototype and more like a real tool.

It's been exhausting. A lot of the work has lived in that deeply unglamorous category of problems that only show up once you think the big ideas are finished: edge cases, formatting, timing, and all the tiny places a writing tool can lose the thread if you're not careful.

But that long final stretch has also been useful. It has given me time to think about how many different ways Betty could be used, and whether keeping her inside a standalone program is actually the right long-term shape.

Why not a Word plugin? Or, maybe even better, a LibreOffice plugin. If part of this project is about giving writers better tools without pushing them deeper into locked platforms, then tech independence should probably be part of the design, not just a slogan taped on afterwards.

I don't have the full answer yet. Right now I'm still making Betty better at the boring but essential parts of editing. But I keep circling back to the idea that she may need to live in more than one place.

— Simon

Why I built Betty.

I write books. I build software. I got tired of the choice between expensive editing tools and sending drafts somewhere I didn't want them to go.

Read more

I built Betty because I wanted a serious editing tool that stayed on my own machine.

I write fiction myself, which means I spend a lot of time in drafts that are half inspired and half held together with stubbornness. I wanted help with the copy and line-editing layer of that work, but I didn't want to upload manuscripts, sign up for another subscription, or build my writing process around somebody else's server.

So I made the thing I wanted to use. Betty is a copy and line editor that runs on your computer, keeps your text local, and gives suggestions without trying to take over the page.

The full name is Bethaniel, but Betty is what stuck. She is opinionated in the useful way, patient about messy drafts, and much faster than I am at spotting the kinds of problems that creep in once you've read the same chapter twelve times.

I am making her free for personal use because I don't think good writing tools should be reserved for people with big budgets. If you're an indie author, a hobbyist, or a small team trying to get words into shape, I want this to be within reach.

I'll use this space to write about what I'm building, what is proving harder than expected, and what I'm learning about writing tools while Betty grows up.

If you'd like to hear when there's a new release or a new post here, there's a mailing list on the main page. I won't use it often, and I won't use it for nonsense.

Until then: you can download Betty from the main page. I hope she helps.

— Simon