Queue 140,000 background jobs somehow #marcbot

Process broadcast messages with lowest priority, otherwise we run into issue where we might broadcast a message to 1,000s of users and when that triggers them to use the bot, their message won't be processed until all the other broadcasts are done sending #marcbot

Create new macOS user account, because the last one somehow broke #life

Get featured on BHG (Better Homes & Gardens) www.bhg.com/ai-and-interior-d… #roomai

Improve submission page design #visiondirectory

Remove like counts because people mistake them for upvotes #betalist

Make prototype vision.directory #visiondirectory

Use more nested fragment caching in todo component #wip

Delete some old (empty) projects that no longer had a user associated with them, because the user deleted their account #wip

Cache logo_color on project records, rather than fetching metadata of logo attachment each time which required 1-2 additional database queries whenever we showed a project's color somewhere #wip

Fix typo that broke homepage #wip

Come up with idea for View Components to natively support caching by running callbacks whenever an associated record is updated #wip (psuedo code)

Use fixed project logo URLs to reduce DB queries. If a project emoji logo doesn't load, please do hard a reload to refresh your browser cache #wip

See response times drop significantly after switching to permanent avatar URLs and fixing indexes for todo similarity search #wip

Move to fixed avatar URLs for all users ( wip.co/@marc/avatar ) and cache it with Cloudflare, so we no longer need to run an ActiveStorage database query each time we show an avatar somewhere #wip

Write long post about finally cracking "social network for makers" #wip twitter.com/marckohlbrugge/st…

Load "recent media" on user profiles in a turbo frame to see if it improves page speed #wip

Add caching keys as HTML comments so I can see if todos are properly cached #wip

Ask author of `neighbors` gem how to optimize my SQL queries, because I just realized that by adding vector embedding to the todos, we're now loading ~25x as much data whenever we're loading todos from the database even though in 99% of cases we don't need the embeddings #wip – github.com/ankane/neighbor/is…

Don't try to load todo comments if comments_count is zero, saving an unnecessary network request and making the site feel a bit more snappy #wip