I want to reflect on my thoughts about how LLMs will be adopted in Enterprise. At the moment, we are seeing the first signs of “hang on - we’re spending a lot here, what are we actually getting from it?”. I first noticed this with the warning from Uber’s COO in May 2026 that it’s becoming harder to justify their spend on it. The current paradigm of allowing the LLM to think and act as an ‘agent’ is what Gary Marcus calls tokenmaxxing. This makes sense as LLM thinking uses up tonnes of output tokens, and allowing them to act autonomously is an even more crazy amount of tokens! ...
Hello World
I’m creating my first blog which I intend to host locally on my mini Topton machine. I’ve found that Hugo seems to be a great static site generator and I’m keen to use it. More soon as I continue to figure this blog, some CICD stuff, and version control._ Update (January 4th, 2025): I did not end up self-hosting this. Instead, I’m using Cloudflare’s Pages and I’ve used their Hugo and Github integrations to automatically rebuild the site when I’ve made changes. I figure (and hope) this setup will save me time in the long-run. ...