<aside> <img src="/icons/help-alternate_gray.svg" alt="/icons/help-alternate_gray.svg" width="40px" />

Why are we doing this:

1. Hiring: people need presense, and things to look at
2. G2M: want people engaged with Pluralis so when we have a run, people will pile on
3. Mindshare and goodwill. Not sure why we need this but know we do.
   1. We want OTHER people to start talking about us.

Three categories:

0. Posts. Blog, explanation, twitter post.
1. Accountments: “we’ve done something. We’ve got X result”
2. Releases. Code, Paper. Something that is useful to others to some extent.

Copy: Get proofreader/formatter → ask Jim.

TO STUDY: exo.

What do we know works:

• Videos showing physical demo of something cool.

• Announcements of Live runs, something working that people thought didn’t

• Really nothing is as impactful as the above - it gets all the attention and matters more than normal marketing slop.

• Huge round annoucements

What doesn’t work/what I don’t want to do

• Countdowns to announcements
• Fake partnership announcements
• Announcements of impending announcements

What message is not there that we have

• DDP pointless. Completely not internalized by market

• No path to ownership. Requires large nodes. Doesn’t scale

• Nous and PI doing DDP.

• Unextractibility so important. Must do

• Vacouso pointless contextless posts. Naval can do it, people I like don’t like it.

May need to ‘smooth brain’ a lot of these announcements.

This whole exercise becomes about execution muscle. High signal + high volumn.

</aside>

Generic

• Single post: “who is Pluralis” team photo. Brief summary “team was at amazon. Alex decided possible, left, everyone joins shortly after. Linkedin and Twitter.
  • Claim blog post did something maybe lol
  • tag max maybe
• All short clips from podcasts that are good. Get someone GOOD for this. Cut up clips, post, repost. Linkedin shorts worth exploring, 0 content on there.
  • must be analytic about it. Who is audience, what’s hitting what’s not. TRACK and adjust.
• Slice up stuff from the Coinfund podcast
• Partnership announcements
  • Oxford.
  • AIML
  • ANU
• Large run collab (simon). Commbank ← this very significant get in some way.

Technical

Sameera

• blog post explaining why errors accumulating make PP hard, and why standard techniques don't work.
• workshop paper: column sparsity in PP
• ICML paper: subspace nets

AJ

• Async PP + forecasting based delay correction. Slightly improves but worse than sync.
  • Understanding Async in PP might be an introductory post but no contribution there.
• Delay correction in the weight space using Nesterov.

Yan + Gil

• A blog post about Swarm framework. User friendly type, like many blog posts that are more accessible to broad audience that doesn’t want to read papers