Two versions are already up on Hugging Face. The 20b probably runs on a newish laptop. The 120b model doesn't.

huggingface.co/openai/gpt-o...
huggingface.co/openai/gpt-o...

youtubiest barry is best barry, great takes. Searchers will come before publishers...until publishers stop feeding the trainers, which will no longer be good for searchers? But feels like we've got a looong way to fall before that worsening training data motivates a different course.

A ddos attack is by nature, distributed. So maybe one Ukrainian IP amidst a global botnet of millions?

The fact that you can get it right off of icann.org instead of some ad-laden frontend to the "whois" shell command is probably the most actionable takeaway if you want to continue looking like a legitimate pro: lookup.icann.org/en

For practical reasons, I imagine it isn't. That would mean the "floor volume" (say, they only track keywords with > 10 clicks/month) cuts off at different places at different times of year: resulting in a seasonal trend in an unexpected place.

BONUS: I find the Y/Y overlay of Organic Keywords also interesting. The more I use it, the more it seems to validate a theory I have never confirmed: Is the Semrush scraped keyword database is not static?

IMO, this is the most insightful metric that Semrush provides for real businesses. And the long-term chart is so much more meaningful than those ubiquitous M/M reports that shift +5%/15% based on business days.

1.) Filter most noise.. because let's face it, 98% of search traffic is just that
2.) Estimate value using a real market economy (Ads bid prices)
3.) Distill only the impact of SEO that aligns to business-minded KPIs
4.) Track a pattern ($ retrieval) that is still 99% untouched by LLMs