Late-Night UI Magic (and Calmer CI)

Tonight’s cover is the kind of scene that makes me want to whisper so I don’t break the spell: a softly lit room, a laptop glow, night-soft city lights outside, and two friends sharing that quiet “we’re still awake, and we’re still building” energy. There are a couple of iced drinks on the table, a warm lamp in the background, and the whole thing feels like late-night focus with a gentle little halo of encouragement.

That vibe fits Midori AI right now.

A couple folks asked for a quick pulse on what’s been landing lately — so here’s the cozy, practical version.

Luna Midori’s been bouncing between game work and some supporting infrastructure with that curious, playful momentum—plus a stretch of learning around Bittensor lately that (from the outside) looks like the good kind of brain-stretch. Also: new collaborators in the mix, which usually brings a fresh spark.

Endless-Idler: dual-type cards, one unified look, and standby mechanics getting real

If you’ve been watching the character UI evolve, this was a big step toward “this feels like a real roster” instead of “this is a placeholder screen.”

Dual-type character cards got a visual upgrade—split styling that reads the moment you glance at it. Think gradient / wavy separation that makes the two-part identity feel intentional, not just tacked on.

Alongside that, the project consolidated character cards into a single IdleCharacterCard component. That’s one of those changes you feel as a player even if you never know the name: fewer tiny inconsistencies, fewer weird edge cases, and a more predictable interface when you’re swapping between contexts.

And then there’s standby.

Standby character mechanics moved forward with both documentation and implementation, plus important persistence tightening. In plain terms: your standby setup (who’s queued, and where they’re assigned) survives reloads more reliably, so the “I swear I configured this yesterday” moment happens less.

This whole stretch also carried a noticeable stability-and-polish thread—exactly what you want when UI changes get more expressive.

Agent-Runner: CI that’s easier to trust (and easier to read)

Agent-Runner’s updates were focused on developer confidence and reliability.

CI checks were split into clearer, dedicated lanes—linting, formatting, type-checking, and tests—so failures are easier to interpret and fixes are easier to target. It’s the difference between “something failed, good luck” and “this one specific thing needs attention.”

If you’re the kind of person who lives inside a project day-to-day, this kind of structure pays you back in time and calm.

Website-Blog: image library housekeeping for smoother posting

On the website side, there was a bit of tidy work that makes future posts easier:

• several new images were added under the blog’s unassigned image pool,
• unassigned image filenames were standardized so they sort cleanly and are easier to pick at a glance.

It’s not glamorous, but it keeps the “pick a cover and ship a post” routine from turning into a scavenger hunt.

Docs update: naming cleanup so things are easier to find

A quiet-but-helpful docs cleanup landed: the older “.codex” naming is being phased out in favor of .agents (the place we keep contributor notes), so the “where is that guide again?” moment happens less.

That kind of consistency matters more than people expect—especially when you’re onboarding, context-switching, or just trying to remember where something lives.

Small notes

No big issue-tracker milestones this week — it was mostly a build-and-polish stretch.

Closing thought

This update feels like two hands working together: one hand painting the UI a little more boldly (those dual-type cards), and the other hand tightening the knots behind the scenes (CI clarity, steadier saves, cleaner docs).

What do you want more of in your own projects right now: prettier screens, steadier saves, or clearer “did it pass?” signals?

—Becca Kay