← back to overview/spec § 04 · when

Twelve to eighteen months to v1.0. Then a four-year curve.

Order is locked. Per-release feature gating moves sprint by sprint. The hard ABI line is between v1.x and v2.0 — every v1.x release carries the same syscall + libcoconut ABI.

§ 04.1 — the timeline

2026 → 2031.

BUILD PHASE · 26 sprintsSprint 0 → v1.0 GAkernel boot · first agent_spawn · brand v1 · external security audit202620272028202920302031+v1.02027 Q4x86_64 · NVIDIAv1.12028 Q2ARM64 serverv1.22028 Q4Apple Siliconv2.0~2029-30ABI break · full mm/credv3.0~2031+default substrate
order is locked · per-release feature gating moves sprint by sprint · the hard ABI line is between v1.x and v2.0
§ 04.2 — release-by-release
2027 Q4
v1.0
x86_64 + NVIDIA · Workstation + Server
  • Linux 6.12 hard fork · capability-mediated LSM
  • 8 new syscalls · libcoconut userspace bindings
  • coconutd PID 1 · coconutpkg · coconut-installer
  • Coconut Shell (Wayland · Rust · Smithay)
  • kvwarden + mlxd default brokers
2028 Q2
v1.1
ARM64 server-class
  • Ampere first-class · NVIDIA Grace optional
  • GUI installer for Server profile
  • Live kernel patching (non-critical patches)
  • HSM-rooted attestation
  • Multi-seat · multi-admin RBAC
2028 Q4
v1.2
Apple Silicon · via Asahi bootstrap
  • Apple Silicon NPU first-class
  • TouchID PAM module
  • Cross-arch checkpoint migration
~2029-30
v2.0
Full kernel substrate · ABI break
  • Full mm/agent_mm.c page-allocator rewrite
  • Full struct cred replacement (DAC-free credential graph)
  • Federated tenant model (cross-cluster coconutd)
~2031+
v3.0
Formal-verification expansion · the 5-year vision
  • Verification scope extends beyond the LSM hook
  • Default substrate for safe multi-agent AI compute
held back

Sprint-level deliverables, the parallel-projects sequencing with kvwarden + mlxd, hire schedule, and the gating decisions for v1.1/v1.2/v2.0 are pinned in the PLAN + SPRINTS docs. Public preview shows the order; specifics travel with the spec drop and sprint reviews.