As chatgpt/better davinci models have levelled the playing field for AI + SEO, I think a better differentiator for AI content will be stuff that really takes advantage of embeddings. A year or two ago you could pretty much whack any niche AI-written stuff on your site and expect it to rank well, because there was such a gold rush to build out content in niche/long-tail content spaces. Now that it's so easy to do this, how do you stay ahead?

One way that's more interesting is in combining embeddings of internal databases with AI-written content.

Just pulled off a bit of an SEO heist using GPT-3, creating 1,800 articles around a competitor brand's main pages by scraping their sitemap. Here's how it went: 1) Find a competitor who ranks for all the sorts of terms you want to rank for, and look at their sitemap. You'll want them to have URLs which are broadly descriptive of what the content at that URL is.

Today was my last day working a 'proper job', as I'm off to focus on freelance work and personal projects for the foreseeable future. Thread below is a collection of thoughts on things we get wrong about 'work', and how I got to this point (1/13) When we're at school/university, we're forced to think about work in a very role-specific way. We get asked "what do you want to be when you're older" where "to be" is interpreted as "what role do you want?" (2/13)

Over the weekend I've been playing around with the Google Trends API, and taking a look at how the rise in interest for Coronavirus has affected what we search for.

1/15: As the panic grows, more & more people are trying to cancel flights 2/15: People are not just trying to claim on existing travel insurance policies, but also to take out new ones: