Roadmap

A rolling engineering roadmap in thematic workstreams across Now / Next / Later. Priorities, not dates.

The GitHub milestones mirror the Now tier and are the source of truth for what's actively in flight; everything below Now is directional and will churn. Canonical copy lives in ROADMAP.md.

Now

The three active milestones — what we're building next.

Engine correctness

The engine must route and terminate exactly as authored — no silent mis-routes, no phantom "Done" on an unresolved gate.

#348 — goal-gate retry re-runs the escalation tail, never the gate, so the pipeline reaches Done with the gate still at outcome=fail.

Epic #308 closeout

Harden build_product against the structural and process gaps surfaced by the case-study runs.

#308 — the epic: structural & process gaps beyond #233.
#304 — budget by cost + a no-progress detector; per-node turn count becomes a backstop.
#307 — document build_product vs superspec, backport the spec-coherence preflight, resolve the examples/ vs workflows/ duplication.

SWE-bench first score

Get a real, published benchmark number.

#465 — first scored SWE-bench Verified run: empty model_patch smoke debug → full Verified run → publish the score.

Directional. Expected to promote to Now as the milestones above close.

Parallel-first resilience

#420 — branch-scoped retry, context, and fix-attempt counters.
#427 — sub-node turn checkpointing for mid-node resume.

Cost & efficiency

#353 — review fan-out cost asymmetry: one reviewer burned 32% of a run duplicating a 42-second finding.

First-run & product polish

#456 — first run fails: build_product hard-exits without SPEC.md; ship a graceful path.
#457 — README information architecture: release-note walls before examples.
#458 — show the TUI: screenshot / GIF in the README and homepage hero.
#459 — positioning: lead with the trust story (budget caps, tamper-evident audit log).

Load-bearing refactors

#396 — replace 11 config-smuggling package globals in cmd/tracker with an explicit runOptions struct.
#393 — claude-code / ACP parsers bypass typed AgentNodeConfig accessors.
#449 — route ~140 raw log.Printf/fmt.Printf diagnostics through a real logger.
#450 / #451 — split the public API from assembly helpers; get suggested_next_nodes off the comma-joined context string.

Later

Backlog. Real, but not scheduled.

Sandbox breadth

#279 — rescope overrideAlreadyRecorded to checkpoint generation.
#280 — file-scoped Bash enforcement for writable_paths.
#281 — per-OS enforcement on macOS (Sandbox) / FreeBSD (Capsicum).

Security documentation & process

#284 — Linux security primitives reference doc.
#285 — 9-class audit checklist for writable_paths changes.
#286 — "freeze and prove" pattern for security PRs.

Structural & cosmetic refactors

#395 — collapse pervasive near-identical duplication (engine emits, llm adapters).
#398 — extract inline prompt: / command: bodies into testable sidecar files.
#452 — write_enriched_sprint.go (1,250 lines) is a domain workflow embedded in agent/.
#453 — split the 1,687-line tracker_doctor.go into unit-testable checks.
#454 — group handler-specific Outcome fields into sub-structs.
#455 — repo hygiene sweep.

Product & positioning

#460 — naming & discoverability: "tracker" is ungoogleable.
#461 — Dippin adoption path: editor support, pipeline gallery.
#463 — run-flag surface (~36 flags) needs presets / progressive disclosure.
#464 — tracker-swebench + tracker-conformance in-repo read as research clutter.

The 1.0 question

#462 — publish a 1.0 roadmap with Go library API stability commitments. Benchmark cadence and the API-stability bar are the gating questions for v1.0.