DEVLOG | April 6, 2026

Native probe runtime

This was a research day: less about ceremony, more about tightening the shape of the thing until it behaved a little more like itself. 11 commits landed or were available locally. There was also a decent amount of local conversation around the work, but the useful part is the trail it leaves: questions asked, decisions narrowed, and a few assumptions made explicit. The center of gravity was server, joystick, damage, hit.

What moved

research moved through tighten pose-adjacent runtime policy gaps, tighten compact transport and status follow-ups, spec dual-client guest probe loop, add phase1 probe hub baseline, add guest probe agent runtime, native xp agent planning, add native probe agent runtime, and add native probe fixture tests.

What I learned

The combat model is becoming less abstract. Hits need to be legible, not just registered, and the visible feedback has to stay tied to the underlying state instead of becoming decorative noise.

Source trail

• Local conversation signal: 4 working sessions.
• research fd3e251: tighten field aggregate damage contract
• research 648f084: tighten remote object control contract
• research 5c3ea10: narrow compact pose controller inputs
• research aa7bac8: tighten pose-adjacent runtime policy gaps
• research 7facd52: tighten compact transport and status follow-ups
• research cd05d4b: spec dual-client guest probe loop
• research 6817fe3: add phase1 probe hub baseline
• research e2cd59f: add guest probe agent runtime
• research 325ab02: native xp agent planning
• research 753786f: add native probe agent runtime
• research 66b8743: add native probe fixture tests

The research lane stays research, and the implementation lane stays implementation. This post is a narrative summary of local work, not a publication of raw originals.