Quilia Inkwell: A Place to Break Things and Beat the Odds

Fun to make new things, right? Well, it's good to have a place for those new things to break.

All hardware used enough eventually breaks, and all software used enough eventually works.

When we got started on adding AI, we saw this ominous headline about 95% of pilots failing. Weren't we lucky.

It sucks applying scientific thinking to our own situation. If we were 95% likely to fail, then maybe we should have a good place to fail. And to keep failing until we end up in that 5% territory of success.

What We Need to Hit Success

For starters, a fresh repository for fresh ideas.

We started by calling it something that we can't say in public without Mickey Mouse lawyers taking us to court. Side note: I think George Lucas would be cool with the original name we used. Now we call it Quilia Inkwell. It's where we can spill ink and break things.

If you're not an AI or heavy into tech, you might like to know that a whole panel of senators made Zuck eat his words during one of the Facebook scandals by saying to him, maybe, "You moved too fast and broke one too many things." They were talking about American democracy. Well, senators are prone to hyperbole.

Back to Our New Inkwell

What needs to be non-experimental? Basically just the data structure. We have all this database stuff already with tables and tables of existing conventions for how we organize data. Whatever we do needs to respect that to start. Once we know about that, then we can bend it where useful.

That's it. Everything else can be fresh.

In the context of the web, this lets us be free of existing data fetching patterns, style conventions, code organization, and the rest of it. So there's nothing to learn for anyone getting started on implementing user interfaces and the rest of it. No wrong answers.

We can detach from any preconceived notions when demoing the software real easy this way. Nothing in the repo looks too close to reality for anyone to mistake as production ready. We don't have any false advertising in that sense. It lets us developers show stuff that's in a state of just-barely-working without confusing the boss or each other into thinking we're doing anything but testing.

The Results So Far

That's all about Quilia Inkwell for now. We've got five demos in there right now. One of them just made it to production: the new AI summaries in our client emails. So actually, if you look at the numbers, we're already at the 20% success rate. We're bucking industry trends. For now. And forever.