hypertwist/docs/v6_5_deep_manual_pack/HyperTwist/AGENTS.md

HyperTwist — AGENTS.md

What HyperTwist is

HyperTwist is the portfolio’s cube and hypercube training environment. It must preserve two identities at once:

• a practical physical-cube recognition, replay, and coaching platform
• a serious higher-dimensional simulation and hypercubing environment

It is not enough for HyperTwist to be good at only one of those.

Authority order

1. repo_portfolio_unified_v6_3.*
2. C:\HyperTwist\docs\HT_REPO_INCORPORATION_AUDIT_2026-05-11.md
3. C:\HyperTwist\docs\HYPERTWIST_REPO_EVALUATION_RESET_AND_IMPLEMENTATION_SCHEDULE_2026-05-11.md
4. C:\HyperTwist\docs\HYPERTWIST_REPO_STATE_BOARD_2026-05-11.md
5. This AGENTS.md
6. The rest of the HyperTwist v6.5 manual set
7. CubeForge initial research as inspiration only, not as code to directly integrate

Core directives

Dual-pillar preservation

Never collapse:

• physical cube pipeline
• hypercube simulation pipeline

They are parallel pillars that reinforce each other.

Unreal-first posture

HyperTwist is native-first. The main runtime should be Unreal Engine 5.4+ with:

• C++ core systems
• Blueprints for UX and training orchestration
• optional OpenXR/VR/MR modes where they materially strengthen higher-dimensional learning

Backbone language bias

Prefer:

• C++
• Rust
• bounded C#

Minimize:

• TypeScript
• Node.js
• Python

But do not remove them where they are still the best practical host for:

• CV/model inference
• certain training or ingestion helpers
• thin companion UIs

Repo classification lens

HyperTwist repos should be judged in these families:

• hypercube simulation foundations
• 3D twisty-puzzle foundations
• physical-cube recognition
• replay and solve analytics
• training and curriculum
• coaching and personalization
• immersive rendering / XR
• export and progression

Coding-model duties

For any HyperTwist repo, the coding model must determine:

• what is truly implemented
• whether it strengthens the physical pipeline, the hyper pipeline, or both
• whether it belongs natively in the runtime, in a sidecar, or only as donor logic
• whether it should be treated as foundation, strategic donor, reserve, or comparator
• whether additional org/index harvesting is still worthwhile

Current implementation truth

For HyperTwist, keep this explicit:

• current shallow-eval set: 71 repos
• currently verified live/implemented in checked Unreal surfaces: 7
• permissive live lanes: 6
• restrictive live lane: 1
• the restrictive landed lane is onionhoney/roux-trainers, and it is to be treated as properly clean-roomed and then implemented

Do not collapse:

• selected
• locked strategic donor
• integrate
• repurpose

into:

• already implemented

Reset rule

Before recommending new HyperTwist donor-shaped widening from the remaining 62 non-live rows:

• do not reopen the already closed Phase 0R, Phase 1R, or Phase 2R packets
• preserve the nine landed rows as current truth
• continue from the closed packet sequence, currently Phase 3R-D