This list is maintained by hand from the POSI adopters page. Last updated: Aug 10, 2026.
What is this?
The POSI site lists the organizations that have adopted the principles, and links to their self-assessments, on its adopters page.
FTR- I created the first version of that list. But I can’t edit it any more.
Adopting POSI is a commitment to keep demonstrating that you follow it, not a one-off announcement. So the useful question isn’t who has adopted the principles — it’s when each of them last checked. The upstream page has never made that easy to see, and since July 2026 it doesn’t publish the dates at all.
This is an attempt to fix that: It lists the same organizations, same links, but sorted around recency, with the number of days since each one’s last self-assessment made obvious.
The list is maintained by hand. The date it was last updated is noted above the list, and the number of days since each organization’s last self-assessment is recalculated every time you load the page.
How did you do it?
It’s a small YAML file. That’s it.
It didn’t start that way. For a couple of years I scraped the POSI adopters page and threw it at an LLM with a JSON schema, because the list was formatted so inconsistently that writing a real parser would have meant a special rule for practically every line. That was one of the rare cases where an unsupervised LLM genuinely was the right tool.
In July 2026 the POSI site restyled the table, with the commit message “restyle adopters table so it stops widening each year”. The table had one column per year, so it did indeed get wider every January. The fix was to bucket everything before 2024 into a single “≤ 2023” column - and to replace every full date with a bare year.
That solved the width problem but it threw away critical detail. “2024” tells you an organization audited itself at some point in a twelve-month window. But you can’t tell if that was three months ago or twenty-one months ago. Which kind of reduces its functionality.
It also broke the scraper. Handed a bare “2024”, the LLM helpfully rounded it to 2024-01-01, so the day counts here were wrong by up to a year with nothing announcing it. That’s the failure mode of non-deterministic scraping with LLMs: it doesn’t stop working, it starts making shit up.
So I stopped deriving the data from a rendered web page. Twenty-nine organizations and forty-three links that change a handful of times a year is nowhere near the threshold where scraping is worth it. I recovered the dates from the last revision of the POSI page that still published them, and I now edit the file by hand.
Using an LLM was fun, but kind of stupid.
Wouldn’t it be better if the POSI site just did this?
Yep.
I’ll make a pull request. If they take it, this page becomes redundant, which would be the best possible outcome for it.