225 - Go Silent

If I go silent then it is usually a sign that things are happening. Our team is one of 8 finalists in a “Startup Showdown” taking place at an AI Conference in SF on September 10th, and the ARC-AGI Challenge will be run with one of our mid-assembly systems soon.

I had previously estimated the probability of being selected as a finalist for such a conference competition as very low, 5%, denoting the default margin of error. It may be a fluke, a true 5% probability event, or it may be that the AI industry is finally starting to look beyond trash technology like LLMs and the various things built on top of them. I can’t even guess the revised probability yet, pending further evidence.

Whether or not the competition gives us a gold star, or whatever one is supposed to “win” at such an event, things are moving in the right direction, and given the right direction then acceleration becomes a “when” rather than an “if”.

The score validation process for ARC-AGI is complicated for anything that can’t simply be uploaded to Kaggle like a proprietary and patent-pending cognitive architecture developed over the past 11 years from scratch, but I’d be rather surprised if they proved unwilling to validate a new high score.

Our CEO will present the company on stage at the 10th anniversary of The AI Conference.

I’ll be grading the competence of the VC panel remotely from Japan.