Anthony Creates Live

Preserving a Brighton notebook

Thu, 10 Sep 2026 00:00:00 GMT

Two weeks of improv intensives in Brighton this summer, and one notebook packed full of my scribbles and jottings on the various warmups, exercises and formats we covered. I was curious to see how LLMs could help me to make sense of those notes and turn them into a permanent record of the experience, and a useful resource for sharing the ideas I'd encountered with others in my improv troupe.

My methodology was pretty simple. I photographed my notebook pages and accompanied them with my spoken recollections of each activity.

My opening prompts:

These are my notes from the musical improv residential. In this chat, I want to think about how I'm gonna make use of them.

I'd like to create a resource which acts as my memory for the week.

That will include the taxonomy, stock of exercises and set of principles that you outline. What I'd like you to do is carefully ask me questions about my notes because I'm pretty sure that quite a lot of them will have to be interpreted or guessed. So I want you to interrogate my notes with a series of questions that I can answer so that you understand what things are referring to and what are just perhaps a memory of a prompt that was used. I'm hoping that by interrogating me we can reconstruct what's most useful and then we'll think about the actual form this takes. Maybe it's a static website, maybe it's some form of dynamic slides.

That got me as far as two canonical markdown files summarising the material, one for week one (musical improv) and one for week two (long-form).

I knew I wanted the two weeks to each have a related but complementary look and feel. Musical improv had felt very silly and sunny, the long-form week a little more serious, and it had coincided with the solar eclipse that week. From my initial chat I created a design brief which I handed over to Claude Design. Here's a summary of how the design progressed for week one:

Stick or Twist: the design journey for the week one page, from brief to page in six moves.

I used GPT Image 2.0 to generate the illustrations which I decided I wanted to accompany the activities.

I then fed my final design along with the canonical content files to Claude Code to build the website. You can explore it here.

Upgrading the Sage

Wed, 09 Sep 2026 00:00:00 GMT

Always Never is my tiny game about inventing proverbs a word at a time. It's inspired by word-at-a-time improv games such as 3-Headed Expert. It's still evolving, but it's live here.

I created it in a week or so in July for OpenAI Build Week. I had to freeze my work on it after that for a while until the competition results were announced (no surprises, I didn't win!).

In recent weeks, I've been working on some improvements.

First, I wanted to give it a visual makeover. I liked the SVG Sage which got created as part of my initial prototype, but I wanted more sophisticated animation.

I got GPT Image to render a series of images of the Sage in different poses to reflect his mood (pleased, frustrated, excited, etc.). I then fed these into Claude Design to work out how to integrate them into the site.

See the full-size mood sheet

Secondly, I really wanted to work on the quality of the Sage's word selection and choices. This is where the game really lives or dies — it has to be fun, and preferably a little wacky. The proverbs that get created need to make sense grammatically, but also have a certain playful, whimsical humour to them. That means improving the selection of words that the Sage offers each round, choosing a good follow-up from the Sage, and better identifying when the proverb might be complete.

This has taken me into interesting terrain. Here's a summary of some things the algorithm is now doing, explained by the model in simple terms ("ELI5, without resorting to inventive metaphors").

From the model

The engine still enforces grammar first: a small state machine tracks which word class can legally come next, so every proverb parses. On top of that, recent work borrows from three fields.

From distributional semantics: every word now carries a compact word embedding derived from ConceptNet Numberbatch, compressed with principal component analysis and shipped with the game. Cosine similarity between these vectors lets the Sage measure how far a candidate word sits from the words already on the table, with no live AI call.

From computational humour: twists are modelled on incongruity-resolution theory, using a local-versus-global surprisal test. A deliberately odd card is only offered if a two-word lookahead finds a way to make it fit, and each twist is labelled with a named rhetorical move such as personification or literalisation.

From search and ranking: word choice uses Gumbel-top-k sampling instead of a fixed argmax, so good words rotate rather than repeat. A regression harness plays hundreds of seeded games, and a Bradley-Terry model ranks policies from players' pairwise votes.

In testing, distinct words offered rose by about a third and hands containing a resolvable twist rose from 86 to 97 percent, with no loss of grammaticality. The next gap is selectional preference, meaning which objects each verb can plausibly take, which is why it can still suggest cracking a rope.

I find it exciting, and a little odd, to be able to develop things like this without needing to master the underlying algorithms and research that they depend on.

What's in a name?

Sat, 25 Jul 2026 00:00:00 GMT

When I registered anthonycreates.com, I was a little disappointed that anthonyerrington.com was already taken. So this was very much a second choice.

However... the name is growing on me. I like creating things. When I came up with the idea for a "Live" corner of my site — a place to document my process, my ideas and my experiments, and to share work in progress as it comes to life — I realised that the choice to register anthonycreates.com feels increasingly apt.

So that's what this space is: Anthony Creates... Live! Work in progress, shared in all its messy process of becoming.