Mark Collier
markcollier.me
Mark Collier
@markcollier.me
2.7K followers 5.7K following 7.6K posts
Austin Powered. OpenStack co-founder, OpenInfra Foundation COO, ex Rackspace & Yahoo! open source for fun & profit. Open Source AI early and often @sparkycollier on twitter and elsewhere Links: markcollier.me
Posts Media Videos Starter Packs
Pinned
Added some more folks to the Open Source AI Starter Pack:

go.bsky.app/N8yVZdW
Today is OpenStack's 15th birthday! What an honor to be part of this community from the beginning. I've met so many friends all over the world and have enjoyed seeing others fall in love with open source and make careers and companies out of it, in OpenStack & beyond! To 15 more!
Reposted by Mark Collier
You're Probably Breaking the Llama Community License

> If you're distributing or redistributing a LLM model that is under the "Llama 3.3 Community License Agreement", you might be breaking at least one of the terms you've explicitly/implicitly agreed to.

notes.victor.earth/youre-probab...
You're Probably Breaking the Llama Community License
You're Probably Breaking the Llama Community License
notes.victor.earth
Who are some cool open source AI folks in Paris I should meet up with?
OpenInfra Summit coming to Paris-Saclay, France Oct 17-19, 2025!

OpenStack, Kata Containers, StarlingX, Zuul + many other open source infrastructure projects will be discussed along with OpenInfra for AI and the mass migration from Vmware to OpenStack

Mark your calendars and practice your French!
💫We are thrilled to share that the #OpenInfraSummit Europe has been officially scheduled for October 17-19, 2025, at École Polytechnique near Paris, France!

Read more about the upcoming event and how you can help build the OpenInfra Summit Europe! openinfra.dev/blog/openinf...
Reposted by Mark Collier
To participate, simply thread one camel through this needle eye
Are you long the $LORD coin?
One of the coolest experiences of my life! And they run a lot of OpenStack to process the insane amounts of data the experiments produce
Are we joy maxing?
Reposted by Mark Collier
Wallace Shawn Emerges As Frontrunner To Replace Daniel Craig As James Bond
or, worse yet, they do understand that and crumbling institutions is the point
Reposted by Mark Collier
OK! My Google colleague Thang Luong shared some exciting updates about AlphaGeometry2!

AG2 now has surpassed the average gold-medalist in solving Olympiad geometry problems, w/ a solve rate of 84% compared to 54% previously!

Paper: arxiv.org/abs/2502.03544
See full list of authors on link
Reposted by Mark Collier
📢 Today we're releasing a new highly detailed dataset for video understanding: HD-EPIC

arxiv.org/abs/2502.04144

hd-epic.github.io

What makes the dataset unique is the vast detail contained in the annotations with 263 annotations per minute over 41 hours of video.
Reposted by Mark Collier
Another seed could be a wikipedia style data curation effort.

In parallel, research data-efficient models that can be trained on 100M tokens. This is comparable to both what humans need to develop speech and to wikipedia, proving that it's possible to curate this amount of data as a community.
First day at FOSDEM and I’m loving it. So much open source and interesting people. Just met a guy who wrote code for F1 to make the cars faster including physics simulation models for designing carbon fiber spoilers and better gear efficiency.
Reposted by Mark Collier
I’d like to learn more about Merriam and Webster. What was that partnership like? I wonder if they ever fought, battled, combatted, wrestled, dueled, or skirmished?
Reposted by Mark Collier
I used the DeepSeek R1 reasoning model to prepare for a new course proposal. These screenshots show with and without the "DeepThink" option turned on--strikingly different. R1 does a lot more synthesis and offers clearer suggestions. It also accepts pdf files, o1 doesn't. Crazy this is open source.
it really tied the room together