LumiGap is a macOS AI vision workspace for poker table recognition.

It reads visible online poker tables from your screen, uses OCR and Core ML to recognize names, stacks, bets, cards, board, pot, and table regions, then turns everything into structured data you can review, correct, export, and use for training custom models.

Build your own poker vision datasets, annotate cards and player tokens, tune recognition thresholds, connect external detectors, convert manifests into model-ready datasets, and test how your models perform on real table layouts.

LumiGap is built for poker researchers, ML experimenters, coaches, and advanced players who want to create their own recognition pipeline instead of relying only on generic trackers or manual screenshots.

It is designed for training, research, dataset creation, and post-session analysis. Users are responsible for following the rules of any poker platforms they use.

CodeHerald provides a new way to keep track of your code review queue, grouped by your next action needed.

When would you use CodeHerald?

• You work in a team that does code reviews.
• Your team receives ad-hoc code review requests via multiple channels: DMs, emails, bookmarks of filtered lists.
• Your team sometimes loses track of small pull requests, delaying them days.
• Your team find ad-hoc code review requests distracting, but cannot put a finger on why.
• Your team tried different strategies to improve code review process, and none of them felt right.

If any of the above is true, CodeHerald will help you.

What can CodeHerald do for you?

CodeHerald groups pull requests by next action: must review, needs an update, can be merged. It allows you to replace slack, emails, filters, and browser bookmarks with one single page that you can open at a glance and decide which PR to tackle next.