The update, and the plan I'm keeping deliberately vague
A real project update on what's working so far, and the shape of the thing we're building next — described in enough detail to be exciting and little enough to stay ours.
Notes from the lead dev.
A working log from Kelly, lead developer at RabbitCodex. Three times a week on what actually happened in AI, and what it changes for the people who have to ship something on Monday.
3 a week · video on Friday · written by Kelly
Friday: we are not starting a newsletter
Every dev log eventually faces the newsletter question. We answered it with a doorbell instead: a tell-me-first list that only rings when there is actually something at the door.
I asked three assistants about us, and graded their answers
Two weeks ago I wrote that search is becoming one answer and the game is being the source that gets cited. This week I ran the experiment on ourselves: three assistants, six questions, one scorecard, and a list of exactly where our writing fails the citation test.
Our homepage told Google to forget it exists. That ends today.
Somewhere along the way our homepage picked up a noindex tag, a one-line meta description, and no social cards at all. This morning I took the tag off, gave the front door a proper introduction, and left structured data on the step like a calling card.
Friday: what we are not saying yet
The weekly update. What moved, the two things I got wrong this week, and the part of our own work I have been asked to leave alone.
Announced is not shipped, and the difference costs weeks
Most coverage treats a blog post and a working product as the same event. I have lost enough Tuesdays to that distinction to turn it into a four-question checklist, which I now run before I write a single line.
Search is turning into one answer, and almost nobody is optimising for it
Ranking tenth on a page of links used to be worth something. When an assistant answers the question instead, there is one source and there is no page. I think this is the most interesting thing happening in my field, and I appear to be almost alone in that.
You can be told first
No newsletter. When there is something to see, the people on this list see it before the log does.
What we are actually doing
RabbitCodex builds and runs business infrastructure using AI agents rather than headcount. Modernisation, data plumbing, integration, and being found by AI assistants rather than by Google alone.
Being the source an AI cites
Answer engine optimisation
Search is becoming answers. When someone asks an assistant a question, one source gets quoted and the rest do not exist. That is a different discipline from ranking tenth on a page of blue links, and almost nobody is doing it deliberately yet.
Edge-first, boring, fast
Infrastructure
Systems that run at the edge, cost little, and do not need a person watching them. Most businesses are paying for servers to sit idle.
Getting off the thing that is holding you back
Modernisation
The spreadsheet that runs the company. The database nobody will touch. The integration that a person does by hand every Tuesday.
Not chatbots. Workers.
Agents that do the work
An agent that drafts, checks, and files is worth more than an agent that chats. The difference is whether it is wired into anything.