Major Architecture Change

Chose to archive the old behavior tree system and rebuild from scratch with Unity's DOTS/ECS architecture. This is a complete structural pivot, not just refactoring.

The Pivot Decision

I pivoted hard today. Yesterday's advice from the AI models was sensible: do not nuke the project, refactor the blackboard, stabilize GC, then migrate. That was the adult path. This morning, after wrestling the old tree again, I chose the other path. I archived the old pile and stood up a fresh DOTS and ECS behavior tree.

The cost was immediate. Twelve hours in, the project has a clean spine but no face. Components exist for intent, systems for execution, and a schedule that should, in theory, run every frame. In practice, an allocation is slipping somewhere and agents are refusing to spawn. It is either a lifetime bug, a disposal miss, or a world mismatch. All of them are fixable, none of them are photogenic.

What Went Right

  • Structural clarity achieved - The new tree reads like a plan instead of a pile
  • Proper separation of concerns - State is data, execution is systems
  • Clean architecture foundation - The "what" and "how" finally live in different places
  • Project identity established - New repo, new name: "Another Run"

What Sucked

  • Complete rebuild cost - Twelve hours in with no visible progress
  • Technical issues - Allocation leaks and agents refusing to spawn
  • No immediate gratification - All the pain now, benefits later
  • Debugging complexity - Could be lifetime bugs, disposal misses, or world mismatches

New Project Identity

Another Run - It fits the loop, it nods to anotherNRE, and it says exactly what the game is about. The subtitle can wait until the world has a mood board and the enemies have real teeth.

Key Decisions

  • Full DOTS/ECS migration instead of incremental refactoring
  • Archive old codebase rather than trying to salvage it
  • New repository and project name - "Another Run"
  • Focus on structural soundness over immediate visual progress

Personal / Meta

Feel less like a random student, more like a tiny game studio. Still broke, still hungry, but at least I own my name now.

Next Up

Tomorrow is not about clever AI. It is about honesty. No leaks. One agent. One tick through the tree. If that happens, the screenshot follows, the demo follows, and the rest of the week is tuning. If it doesn't, I keep going until it does.

Tomorrow's Success Criteria

One agent successfully running through one complete behavior tree tick without memory leaks. Everything else is secondary.

Technical Reflection

The win is structural clarity. The new tree reads like a plan instead of a pile. State is data. Execution is systems. The "what" and the "how" finally live in different places. That is the point of this rewrite. It is also why it hurts now and pays later.

This is the brave path - sometimes you have to burn the old structure to build something truly better. The adult path would have been safer, but this path has the potential for something genuinely excellent.