renames

CONTEXT.md

@@ -7,7 +7,7 @@ This context defines the core workflow runtime language used by `wf_core`,
 
 **Scheduler Foundation**:
 The runtime model that selects runnable frames and advances workflow execution without assuming there is only one active cursor.
-_Avoid_: Concurrent foreach, parallel foreach foundation
+_Avoid_: Foreach feature, concurrent foreach implementation
 
 **Frame**:
 An execution cursor for one active portion of a workflow run.
@@ -26,7 +26,7 @@ The ordered list of runnable frame identifiers that defines deterministic schedu
 _Avoid_: Frame scan, implicit dict order
 
 **Foreach Policy**:
-The future runtime configuration that decides parallel foreach batching, item failure handling, and cancellation/drain behavior.
+The future runtime configuration that decides concurrent foreach admission, item failure handling, and quiescence behavior.
 _Avoid_: Parallel flag
 
 **Blocked Frame**:
@@ -62,7 +62,7 @@ The committed state a frame may read based on workflow topology and completed up
 _Avoid_: Shared live object
 
 **Lineage Isolation**:
-The rule that a parallel child frame sees parent-visible state plus its own ancestor writes, but not sibling branch writes.
+The rule that a concurrent child frame sees parent-visible state plus its own ancestor writes, but not sibling branch writes.
 _Avoid_: Global committed state
 
 **Barrier**:
@@ -92,15 +92,15 @@ _Avoid_: Job, invocation
   the child/event it waits on wakes it.
 - A **Blocked Frame** should carry a **Block Reason** so deadlocks and wakeups
   are explainable.
-- Parallel foreach and native subgraphs depend on the **Scheduler Foundation**.
+- Concurrent foreach and native subgraphs depend on the **Scheduler Foundation**.
-- Parallel foreach requires a **Foreach Policy** before public support is
+- Concurrent foreach requires a **Foreach Policy** before public support is
   enabled.
-- Future parallel foreach should be bounded by default. `max_active` and
+- Future concurrent foreach should be bounded by default. `max_active` and
   `max_outstanding` prevent one workflow from spawning unbounded MCP, HTTP,
   browser, or external service calls.
-- Future concurrency-specific foreach settings should live in a nested parallel
+- Future concurrency-specific foreach settings should live in a nested
   policy object rather than expanding `ForeachNode` with many top-level fields.
-- Item error handling is foreach-wide, not parallel-only. Serial and parallel
+- Item error handling is foreach-wide, not concurrent-only. Serial and concurrent
   foreach can both profit from `fail`, `skip`, or `collect` item failure policy.
 - `collect` item error policy must declare an explicit destination for
   structured item errors. Collected errors should be ordered by item index, not
@@ -137,7 +137,7 @@ _Avoid_: Job, invocation
   failure was collected.
 - `skip` emits `completed_with_errors` when one or more item failures were
   skipped. It emits `done` only when all items succeed.
-- Parallel `fail` item policy should stop scheduling new items, drain already
+- Concurrent `fail` item policy should stop scheduling new items, drain already
   started jobs to a quiescent point, capture their results safely, and then fail
   the run. It should not assume hard cancellation is safe.
 - After `fail` trips, drained sibling results are for trace/observability and
@@ -145,7 +145,7 @@ _Avoid_: Job, invocation
   boundary.
 - Foreach modes that continue after item failure, such as `collect` and `skip`,
   should buffer item state patches until the foreach barrier completes. Future
-  parallel foreach should always use barrier-buffered commits.
+  concurrent foreach should always use barrier-buffered commits.
 - Serial `fail` may keep immediate commits for compatibility. Commit strategy
   should be extracted into runtime helpers instead of being smeared through
   foreach execution code.
@@ -161,7 +161,7 @@ _Avoid_: Job, invocation
 - Pending barrier results must live in resumable `RunState`/frame metadata, not
   only in trace. Trace records history; runtime state is what resume/checkpoint
   uses.
-- Parallel item frames need lineage-local pending state: later nodes in the same
+- Concurrent item frames need lineage-local pending state: later nodes in the same
   item lineage can read earlier pending patches from that item, while sibling
   items cannot. The exact aggregate output API for committing item results to
   parent state is deferred.
@@ -224,10 +224,10 @@ _Avoid_: Job, invocation
   but structured foreach refs should be preferred for non-trivial authoring.
 - Future foreach policy shape should validate cross-field rules: `collect`
   requires `collect_to`, non-collect actions forbid `collect_to`,
-  `mode="parallel"` requires a parallel policy, and `mode="serial"` forbids a
+  `mode="concurrent"` requires a concurrent policy, and `mode="serial"` forbids
-  parallel policy. Deprecated top-level `on_item_error` may parse into the
+  a concurrent policy. Deprecated top-level `on_item_error` may parse into the
   nested item error policy, but canonical dumps should use the nested shape.
-- `ForeachParallelPolicy` should split limits into `max_active` and
+- `ForeachConcurrentPolicy` should split limits into `max_active` and
   `max_outstanding`. Defaults are `max_active=4` and `max_outstanding=20`;
   validation requires `max_outstanding >= max_active`. Ready or running item
   frames consume active capacity; blocked item frames consume outstanding
@@ -260,7 +260,7 @@ _Avoid_: Job, invocation
   backpressure inside the admitted node-call boundary, not workflow-level
   `BLOCKED` state that frees foreach active capacity.
 - An **Interrupt** pauses the whole **Run**, even if the interrupted frame is a
-  child of future parallel work.
+  child of future concurrent work.
 - A **Runtime Failure** is distinct from a node returning an `error` outcome.
   Outcomes are graph control flow; runtime failures are scheduler stops unless
   a policy handles them.
@@ -269,32 +269,34 @@ _Avoid_: Job, invocation
   **Runtime Failure**.
 - Missing edges for declared outcomes are validation errors, not normal runtime
   branch policy.
-- Unsupported parallel foreach semantics should be rejected by validation before
+- Unsupported concurrent foreach semantics should be rejected by validation before
   runtime. Runtime may stay defensive, but validation owns the user-facing gate.
 - `on_item_error="collect"` and `"skip"` are future policy shapes unless runtime
   support is explicitly implemented. Current scheduler work should make official
   support easier, not pretend it already exists.
 - A **Trace** records actual scheduler execution order; grouping or sorting by
   foreach index is a presentation concern.
-- Parallel child frames may write to the same state path only through a
+- Concurrent child frames may write to the same state path only through a
   **Reducer**.
-- Future parallel frames should produce **State Patches** that the scheduler
+- Future concurrent frames should produce **State Patches** that the scheduler
   commits atomically.
 - A frame's **State Visibility** excludes uncommitted sibling writes.
-- Future parallel branches require **Lineage Isolation**: sibling branch writes
+- Future concurrent branches require **Lineage Isolation**: sibling branch writes
   are invisible unless the graph explicitly joins or merges them.
-- A **Barrier** is the explicit merge boundary for parallel lineage patches.
+- A **Barrier** is the explicit merge boundary for concurrent lineage patches.
 - Foreach may own an implicit **Barrier**; future graph-level convergence may use
   an explicit barrier node.
 
 ## Example Dialogue
 
-> **Dev:** "Are we implementing parallel foreach now?"
+> **Dev:** "Are we implementing concurrent foreach now?"
 > **Domain expert:** "No. First we are implementing the **Scheduler Foundation** so a **Run** can eventually manage multiple runnable **Frames** safely."
 
 ## Flagged Ambiguities
 
-- "concurrent foreach" was used for the next task, but the resolved scope is **Scheduler Foundation**: deterministic internal runtime prep before public parallel foreach support.
+- "concurrent foreach" is the future workflow mode. The resolved current scope is
+  **Scheduler Foundation**: deterministic internal runtime prep before public
+  concurrent foreach support.
 - `current_frame_id` / `current_node_id` are compatibility fields for the selected cursor, not the source of truth for all runnable work.
 - `current_frame_id` remains persisted for compatibility, but means "the frame
   currently selected by the scheduler", not "the only live frame".
@@ -311,7 +313,7 @@ _Avoid_: Job, invocation
   interrupt reason, because it carries externally visible resume semantics.
 - Scheduling order should be explicit through a **Ready Queue**, not inferred
   from frame dictionary iteration.
-- `mode="parallel"` stays unsupported until **Foreach Policy** and parent
+- `mode="concurrent"` stays unsupported until **Foreach Policy** and parent
   completion semantics are explicit.
 - A **Run** has at most one outstanding **Interrupt**. Resume wakes the frame
   referenced by that interrupt and scheduling continues from the **Ready Queue**.
@@ -321,15 +323,16 @@ _Avoid_: Job, invocation
 - Ready sibling frames are preserved during an **Interrupt**, but the resumed
   frame is placed at the front of the **Ready Queue** before scheduling
   continues.
-- Future parallel execution should not assume in-flight node calls can be
+- Future concurrent execution should not assume in-flight node calls can be
   safely cancelled. When an **Interrupt** occurs, scheduling should stop; already
   started jobs should drain to pending results before resume/commit policy
   decides what becomes visible.
-- Future parallel interrupts should return control to the caller only at a
+- Future concurrent interrupts should return control to the caller only at a
   quiescent pause point: no new work is scheduled, already-started jobs have
   drained, and their results are captured without unsafe commits.
 - Future async execution must protect append-only **Trace** writes, but should
-  not change trace semantics.
+  not change trace semantics. Async runtime is an execution capability for
+  simultaneous async node handlers, not a separate foreach workflow mode.
 - Scheduler block/wake events should not be added to the public **Trace** in
   the first pass. Future observability, including OpenTelemetry-style spans, is
   a separate concern.
@@ -339,7 +342,7 @@ _Avoid_: Job, invocation
   but raise for malformed metadata on the right kind. Corrupt runtime metadata
   is a runtime invariant failure.
 - First-pass **Block Reason** support only needs child-frame blocking; future
-  interrupt, subgraph, and parallel barrier reasons can extend the same shape.
+  interrupt, subgraph, and concurrent barrier reasons can extend the same shape.
 - Frame creation must reject duplicate frame identifiers. The **Ready Queue**
   must contain only existing pending frames and must never contain duplicate
   frame identifiers.
@@ -378,10 +381,10 @@ _Avoid_: Job, invocation
 - Serial foreach blocks the parent on one iteration child at a time, so the
   child runs until completion/interruption/failure before the parent wakes.
 - Future async execution must protect shared state writes. Last-writer-wins is
-  not an acceptable merge policy for parallel child frames.
+  not an acceptable merge policy for concurrent child frames.
-- Serial execution may continue mutating state immediately until parallel
+- Serial execution may continue mutating state immediately until concurrent
   execution needs scheduler-controlled **State Patch** commits.
-- Parallel sibling frames must not depend on observing each other's writes. Use
+- Concurrent sibling frames must not depend on observing each other's writes. Use
   serial foreach or explicit graph structure for ordered dependencies.
 - Missing reducers mean replace only within a single serial lineage. At a
   **Barrier**, multiple writes to the same path without a reducer are conflicts,
@@ -390,10 +393,10 @@ _Avoid_: Job, invocation
   repurposed later only through an explicit design pass.
 - The first **Scheduler Foundation** implementation should create ready/block/wake
   seams only. **Lineage Isolation** and **Barrier** merge behavior are future
-  parallel semantics, not first-pass behavior.
+  concurrent semantics, not first-pass behavior.
 - First-pass scheduler code should add durable seams such as ready queue,
   block/wake helpers, and typed foreach metadata. It should not add public
-  parallel policy, barrier, snapshot, or lineage-patch fields before those
+  concurrent policy, barrier, snapshot, or lineage-patch fields before those
   semantics are enforced.
 - Scheduler rules are shared by sync and async runtime paths; only node handler
   execution differs.

docs/README.md

@@ -81,6 +81,7 @@ Two runtime features may come back after the platform/DX roadmap advances:
 
 - Native subgraph execution with child run state, interrupt bubbling, and resume
   back into the child workflow.
-- Async parallel foreach with explicit scheduling, reducer/merge semantics, and
+- Concurrent foreach with explicit scheduling, reducer/merge semantics, and
-  failure policy. Do not implement this as plain `asyncio.gather` over sync
+  failure policy. Sync runtime can interleave item frames deterministically;
-  handlers.
+  async runtime can add simultaneous async node handler execution. Do not
+  implement this as plain `asyncio.gather` over sync handlers.

docs/adr/0001-scheduler-foundation-before-concurrent-foreach.md

@@ -1,10 +1,10 @@
-# Scheduler Foundation Before Parallel Foreach
+# Scheduler Foundation Before Concurrent Foreach
 
 Status: accepted
 
 We will introduce an explicit scheduler foundation in `wf_core` before enabling
-parallel foreach. The first implementation preserves deterministic serial
+concurrent foreach. The first implementation preserves deterministic serial
-behavior while adding ready/block/wake semantics, because parallel foreach and
+behavior while adding ready/block/wake semantics, because concurrent foreach and
 native subgraphs both require multiple runnable frames without treating
 `current_frame_id` as the whole runtime state.
 
@@ -15,9 +15,9 @@ The current runtime behaves like a stack cursor. `RunState.current_frame_id` and
 at a time before control collapses back to the parent. This works for serial
 execution, but it does not give the runtime a clear way to represent multiple
 runnable frames, blocked parent frames, interrupt resume priority, or future
-parallel branch merge semantics.
+concurrent branch merge semantics.
 
-Parallel foreach is not just `asyncio.gather` over node calls. It changes state
+Concurrent foreach is not just `asyncio.gather` over node calls. It changes state
 visibility, error handling, interrupt behavior, trace ordering, reducer rules,
 and parent completion semantics. Enabling it before those concepts exist would
 make the runtime fragile.
@@ -25,7 +25,7 @@ make the runtime fragile.
 ## Decision
 
 Add an internal scheduler foundation first. The first pass will keep serial
-workflow behavior intact and will not enable `foreach(mode="parallel")`.
+workflow behavior intact and will not enable `foreach(mode="concurrent")`.
 
 The scheduler model is:
 
@@ -56,7 +56,7 @@ the next item or finish.
 
 ## Rejected Alternatives
 
-**Patch parallelism directly into foreach.**
+**Patch concurrent execution directly into foreach.**
 This would keep the stack-cursor model and force foreach to own scheduling,
 interrupt, merge, and failure policy. It would not help native subgraphs.
 
@@ -74,22 +74,22 @@ refine this later.
 The future barrier semantics are larger than the current join node. `JoinNode`
 may be repurposed later, but not silently in this scheduler pass.
 
-## Future Parallel Semantics
+## Future Concurrent Semantics
 
-Parallel foreach and graph-level convergence require additional semantics before
+Concurrent foreach and graph-level convergence require additional semantics before
 they are enabled:
 
-- A future foreach policy must define batching, `max_concurrency`, item failure
+- A future foreach policy must define admission limits, item failure handling,
-  handling, and cancellation/drain behavior.
+  and quiescence behavior.
-- Parallel foreach should be bounded by default so workflows do not spawn
+- Concurrent foreach should be bounded by default so workflows do not spawn
   unbounded MCP, HTTP, browser, or external service calls.
 - A future barrier must wait for multiple child or upstream frames and merge
   their lineage patches.
 - Missing reducers mean replace only within one serial lineage. At a barrier,
   multiple writes to the same path without a reducer are conflicts.
-- Future parallel child frames should produce state patches that the scheduler
+- Future concurrent child frames should produce state patches that the scheduler
   commits atomically.
-- Parallel branches require lineage isolation: a child frame sees parent-visible
+- Concurrent branches require lineage isolation: a child frame sees parent-visible
   state plus its own ancestor writes, not sibling branch writes.
 - Public trace remains append-only chronological execution history. Scheduler
   block/wake events are not added to public trace in this pass.
@@ -104,7 +104,7 @@ Resume wakes the interrupted frame, clears the run interrupt, and places the
 resumed frame at the front of the ready queue. Ready sibling frames are preserved
 but do not run while the interrupt is outstanding.
 
-Future parallel execution should not assume in-flight node calls can be safely
+Future concurrent execution should not assume in-flight node calls can be safely
 cancelled. If one frame interrupts while sibling jobs are already started, the
 runtime should stop scheduling new work, let started jobs drain to pending
 results, and defer sibling state commits until resume/commit policy decides what
@@ -120,7 +120,7 @@ are graph control flow. Runtime failures stop scheduling unless a future policy
 explicitly handles them.
 
 In the first scheduler pass, any runtime failure fails the whole run. Future
-parallel foreach policies may support `collect` or `skip`, but those are policy
+concurrent foreach policies may support `collect` or `skip`, but those are policy
 features and should not be treated as implemented until runtime support exists.
 
 ## Compatibility Requirements

docs/adr/0002-concurrent-foreach-policy-and-barrier-commits.md

@@ -1,36 +1,40 @@
-# Parallel Foreach Policy and Barrier Commits
+# Concurrent Foreach Policy and Barrier Commits
 
 Status: accepted
 
-Parallel foreach will use bounded scheduling, lineage-local pending state, and
+Concurrent foreach will use bounded scheduling, lineage-local pending state, and
-barrier-buffered commits. This keeps parallel item execution deterministic,
+barrier-buffered commits. This keeps item execution deterministic, resumable,
-resumable, and safe around reducers, interrupts, and external tool calls.
+and safe around reducers, interrupts, and external tool calls.
 
 ## Context
 
-The scheduler foundation makes multiple runnable frames possible, but parallel
+The scheduler foundation makes multiple runnable frames possible, but concurrent
-foreach needs more than concurrent node calls. Each item lineage may execute
+foreach needs more than multiple node calls. Each item lineage may execute
 several nodes, read its own pending writes, fail independently, block, or
 interrupt the whole run. Sibling item writes must not leak into each other, and
 state commits need deterministic merge behavior.
 
 ## Decision
 
-Parallel foreach will be async-runtime-only and bounded by policy.
+Concurrent foreach is a workflow mode and must be bounded by policy.
 
 The future foreach model should separate item error behavior from concurrency
 behavior:
 
-- `item_error` is foreach-wide and applies to serial and parallel foreach.
+- `item_error` is foreach-wide and applies to serial and concurrent foreach.
-- `parallel` is a nested policy object used only when `mode="parallel"`.
+- `concurrent` is a nested policy object used only when
-- `parallel.max_active` defaults to `4`.
+  `mode="concurrent"`.
-- `parallel.max_outstanding` defaults to `20`.
+- `concurrent.max_active` defaults to `4`.
+- `concurrent.max_outstanding` defaults to `20`.
 - Validation requires `max_outstanding >= max_active`.
 - Ready or running item frames consume active capacity.
 - Blocked item frames consume outstanding capacity but not active capacity.
 
-`foreach(mode="parallel")` requires async execution. Sync execution should reject
+`foreach(mode="concurrent")` should be executable by the sync runtime as
-it clearly rather than inventing thread/process semantics.
+deterministic frame interleaving: one admitted node handler call at a time. The
+async runtime may additionally run admitted async node handler calls
+simultaneously. Sync handlers should not be pushed into thread/process
+parallelism by default.
 
 ## Item Error Policy
 
@@ -71,7 +75,7 @@ failure as control flow, the outcome should be `failed`.
 
 ## Barrier-Buffered Commits
 
-Parallel foreach item writes are buffered as pending state patches/results, not
+Concurrent foreach item writes are buffered as pending state patches/results, not
 committed directly to `RunState.state`.
 
 The state model is:
@@ -88,8 +92,8 @@ Completion-order merging may exist later as an explicit barrier merge policy,
 not reducer behavior.
 
 Patch creation and commit must extract/reuse the existing node output
-validation, output binding, and reducer logic. Parallel foreach must not create
+validation, output binding, and reducer logic. Concurrent foreach must not
-a second write system.
+create a second write system.
 
 ## Merge and Reducer Rules
 
@@ -103,7 +107,7 @@ means item index order.
 
 ## Interrupt and Failure Quiescence
 
-Future parallel execution should not assume in-flight node calls can be safely
+Future concurrent execution should not assume in-flight node calls can be safely
 cancelled.
 
 If an interrupt or fail policy trips while sibling jobs are already started, the
@@ -131,7 +135,7 @@ node calls, and those waits keep the frame `RUNNING` rather than `BLOCKED`.
 
 ## Context and Lineage
 
-Parallel item frames need lineage-local pending state. Later nodes in the same
+Concurrent item frames need lineage-local pending state. Later nodes in the same
 item lineage can read earlier pending patches from that item, while siblings
 cannot.
 
@@ -157,5 +161,5 @@ Forks produce branch lineage tokens; gathers consume declared token sets and
 produce merged tokens. This supports partial gathers such as merging `a+b`
 before later merging with `c`.
 
-Parallel foreach remains the nearer target because its implicit lineage tokens
+Concurrent foreach remains the nearer target because its implicit lineage tokens
 are item indexes owned by one foreach activation.

docs/current_roadmap.md

@@ -40,15 +40,17 @@ implementation state.
 ## Runtime and Platform Roadmap
 
 - Scheduler foundation decision record:
-  [ADR 0001](./adr/0001-scheduler-foundation-before-parallel-foreach.md).
+  [ADR 0001](./adr/0001-scheduler-foundation-before-concurrent-foreach.md).
-- Parallel foreach policy decision record:
+- Concurrent foreach policy decision record:
-  [ADR 0002](./adr/0002-parallel-foreach-policy-and-barrier-commits.md).
+  [ADR 0002](./adr/0002-concurrent-foreach-policy-and-barrier-commits.md).
 - **Native subgraphs / graph-as-node**: add child run state, child trace
   preservation, interrupt bubbling, and resume back into the child workflow.
   Wrapper artifacts currently execute as deployments and return run status;
   true graph-as-node outcome propagation belongs here.
-- **Async parallel foreach**: add explicit scheduling, reducer/merge semantics,
+- **Concurrent foreach**: add explicit scheduling, reducer/merge semantics, and
-  and failure policy. Do not model this as plain parallel calls over sync
+  failure policy. Sync runtime can interleave admitted item frames one node at a
+  time; async runtime can additionally run admitted async node handlers
+  simultaneously. Do not model this as plain `asyncio.gather` over sync
   handlers.
 - **Persistent run history**: add a run store before adding stable `run_id`,
   `inspect_run`, or `read_run_trace(run_id, range)` APIs. Current traces are
@@ -64,10 +66,10 @@ implementation state.
 
 Frame stress points to solve before either feature:
 
-- `RunState.current_frame_id` currently models one active execution cursor.
+- `RunState.current_frame_id` currently models the selected execution cursor.
-  Parallel foreach likely needs multiple runnable child frames.
+  Concurrent foreach needs multiple runnable child frames.
 - `ExecutionFrame.metadata` currently carries ad hoc foreach data. Subgraphs and
-  parallel foreach should get typed frame payloads or strongly bounded helper
+  concurrent foreach should get typed frame payloads or strongly bounded helper
   accessors before metadata grows more meanings.
 - Subgraph frames need child workflow identity/version/deployment binding, not
   just a generic metadata dictionary.
@@ -79,6 +81,6 @@ Frame stress points to solve before either feature:
 
 The MCP workflow authoring path is now usable enough for real testing. The next
 bottleneck is runtime/platform correctness: resumable child execution,
-parallel scheduling, persistent run history, and protocol-native progress
+concurrent scheduling, persistent run history, and protocol-native progress
 reporting. Those pieces should come before adding more high-level authoring
 sugar.

docs/superpowers/plans/2026-05-22-concurrent-foreach-roadmap.md

@@ -1,10 +1,10 @@
-# Parallel Foreach Roadmap Implementation Plan
+# Concurrent Foreach Roadmap Implementation Plan
 
 > **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
 
-**Goal:** Implement parallel foreach incrementally without breaking serial workflows or duplicating state-write logic.
+**Goal:** Implement concurrent foreach incrementally without breaking serial workflows or duplicating state-write logic.
 
-**Architecture:** The work is split into four independently shippable layers: policy models, state patch extraction, barrier runtime state, and async parallel execution. Each layer preserves current serial behavior and adds tests before implementation. `foreach(mode="parallel")` remains unsupported until the final layer.
+**Architecture:** The work is split into four independently shippable layers: policy models, state patch extraction, barrier runtime state, and concurrent execution. Each layer preserves current serial behavior and adds tests before implementation. `foreach(mode="concurrent")` remains unsupported until the final layer. Sync runtime should support deterministic interleaving once the mode is enabled; async runtime can additionally run admitted async node handlers simultaneously.
 
 **Tech Stack:** Python 3.14, Pydantic v2, dataclasses, pytest, basedpyright, ruff, existing `wf_core` scheduler/runtime modules.
 
@@ -48,7 +48,7 @@ def test_serial_foreach_defaults_to_fail_item_policy() -> None:
     assert node.mode == "serial"
     assert node.item_error.action == "fail"
     assert node.item_error.collect_to is None
-    assert node.parallel is None
+    assert node.concurrent is None
 
 
 def test_collect_item_policy_requires_collect_to() -> None:
@@ -64,20 +64,20 @@ def test_collect_item_policy_requires_collect_to() -> None:
         )
 
 
-def test_parallel_policy_requires_parallel_mode() -> None:
+def test_concurrent_policy_requires_concurrent_mode() -> None:
-    with pytest.raises(ValidationError, match="parallel policy"):
+    with pytest.raises(ValidationError, match="concurrent policy"):
         ForeachNode.model_validate(
             {
                 "id": "each",
                 "type": "foreach",
                 "over": {"root": "state", "parts": ["items"]},
                 "as": "item",
-                "parallel": {"max_active": 4, "max_outstanding": 20},
+                "concurrent": {"max_active": 4, "max_outstanding": 20},
             }
         )
 
 
-def test_parallel_policy_validates_capacity_order() -> None:
+def test_concurrent_policy_validates_capacity_order() -> None:
     with pytest.raises(ValidationError, match="max_outstanding"):
         ForeachNode.model_validate(
             {
@@ -85,8 +85,8 @@ def test_parallel_policy_validates_capacity_order() -> None:
                 "type": "foreach",
                 "over": {"root": "state", "parts": ["items"]},
                 "as": "item",
-                "mode": "parallel",
+                "mode": "concurrent",
-                "parallel": {"max_active": 10, "max_outstanding": 4},
+                "concurrent": {"max_active": 10, "max_outstanding": 4},
             }
         )
 ```
@@ -116,8 +116,8 @@ class ForeachItemErrorPolicy(BaseModel):
         return self
 
 
-class ForeachParallelPolicy(BaseModel):
+class ForeachConcurrentPolicy(BaseModel):
-    """Concurrency policy for async parallel foreach execution."""
+    """Concurrency policy for foreach frame admission."""
 
     model_config = ConfigDict(extra="forbid")
 
@@ -136,7 +136,7 @@ Update `ForeachNode`:
 
 ```python
 item_error: ForeachItemErrorPolicy = Field(default_factory=ForeachItemErrorPolicy)
-parallel: ForeachParallelPolicy | None = None
+concurrent: ForeachConcurrentPolicy | None = None
 on_item_error: Literal["fail", "collect", "skip"] | None = Field(
     default=None,
     exclude=True,
@@ -149,10 +149,10 @@ Add a `model_validator(mode="before")` that converts old `on_item_error` into `i
 Add a `model_validator(mode="after")` that enforces:
 
 ```python
-if self.mode == "parallel" and self.parallel is None:
+if self.mode == "concurrent" and self.concurrent is None:
-    raise ValueError("parallel foreach requires parallel policy")
+    raise ValueError("concurrent foreach requires concurrent policy")
-if self.mode == "serial" and self.parallel is not None:
+if self.mode == "serial" and self.concurrent is not None:
-    raise ValueError("parallel policy is only valid when mode='parallel'")
+    raise ValueError("concurrent policy is only valid when mode='concurrent'")
 ```
 
 - [ ] **Step 3: Update derived outcomes**
@@ -185,14 +185,14 @@ In `src/wf_core/runtime/ops/foreach.py`, keep:
 
 ```python
 if step.mode != "serial":
-    raise WorkflowExecutionError("parallel foreach execution is not implemented yet")
+    raise WorkflowExecutionError("concurrent foreach execution is not implemented yet")
 ```
 
 Add a comment:
 
 ```python
 # Policy models are accepted before execution support so saved workflows can
-# validate shape, but runtime must reject parallel until barrier commits exist.
+# validate shape, but runtime must reject concurrent until barrier commits exist.
 ```
 
 - [ ] **Step 6: Verify phase**
@@ -316,7 +316,7 @@ Expected: all pass; full suite should still pass before moving on.
 
 ## Phase 3: Foreach Barrier Runtime State
 
-**Goal:** Add resumable barrier metadata and pending result structures without enabling async parallel execution.
+**Goal:** Add resumable barrier metadata and pending result structures without enabling concurrent execution.
 
 **Files:**
 - Modify: `src/wf_core/runtime/scheduler.py`
@@ -422,9 +422,9 @@ Expected: pass.
 
 ---
 
-## Phase 4: Async Parallel Foreach
+## Phase 4: Concurrent Foreach Execution
 
-**Goal:** Enable `foreach(mode="parallel")` in async execution only, using policy limits, pending results, and barrier commits.
+**Goal:** Enable `foreach(mode="concurrent")` using policy limits, pending results, and barrier commits. Sync runtime interleaves admitted item frames one node call at a time; async runtime may run admitted async node handler calls simultaneously.
 
 **Files:**
 - Modify: `src/wf_core/runtime/ops/foreach.py`
@@ -432,22 +432,22 @@ Expected: pass.
 - Modify: `src/wf_core/runtime/engine.py`
 - Modify: `src/wf_core/runtime/ops/nodes.py`
 - Modify: `src/wf_core/runtime/foreach_state.py`
-- Test: `tests/core/test_parallel_foreach.py`
+- Test: `tests/core/test_concurrent_foreach.py`
 
-- [ ] **Step 1: Add async-only rejection tests**
+- [ ] **Step 1: Add sync interleaving and async execution tests**
 
-Create `tests/core/test_parallel_foreach.py` with:
+Create `tests/core/test_concurrent_foreach.py` with:
 
 ```python
-def test_sync_runtime_rejects_parallel_foreach() -> None:
+def test_sync_runtime_interleaves_concurrent_foreach() -> None:
     ...
 
 
-async def test_async_runtime_accepts_parallel_foreach() -> None:
+async def test_async_runtime_accepts_concurrent_foreach() -> None:
     ...
 ```
 
-Expected before implementation: async test fails because runtime still rejects parallel.
+Expected before implementation: both tests fail because runtime still rejects concurrent mode.
 
 - [ ] **Step 2: Add capacity tests**
 
@@ -456,7 +456,7 @@ Use async node handlers that record start/completion order and block on `asyncio
 Test:
 
 ```python
-async def test_parallel_foreach_respects_max_active() -> None:
+async def test_concurrent_foreach_respects_max_active() -> None:
     ...
     assert max_seen_active == 2
 ```
@@ -481,11 +481,11 @@ This may require a small test-only node that blocks through the runtime-supporte
 Tests:
 
 ```python
-async def test_parallel_collect_writes_ordered_errors_and_emits_completed_with_errors() -> None:
+async def test_concurrent_collect_writes_ordered_errors_and_emits_completed_with_errors() -> None:
     ...
 
 
-async def test_parallel_skip_emits_completed_with_errors_without_hidden_state() -> None:
+async def test_concurrent_skip_emits_completed_with_errors_without_hidden_state() -> None:
     ...
 ```
 
@@ -502,17 +502,17 @@ Use nodes that complete out of order but write list-like results.
 
 Assert committed state is ordered by item index, not completion order.
 
-- [ ] **Step 6: Implement async parallel child scheduling**
+- [ ] **Step 6: Implement concurrent child scheduling**
 
 In `step_foreach`, branch by mode:
 
 ```python
 if step.mode == "serial":
     return step_foreach_serial(...)
-return step_foreach_parallel(...)
+return step_foreach_concurrent(...)
 ```
 
-`step_foreach_parallel` should:
+`step_foreach_concurrent` should:
 - inspect `ForeachBarrierState`
 - start children while `active < max_active` and `outstanding < max_outstanding`
 - block parent when waiting for children
@@ -530,14 +530,14 @@ class RuntimeLimits:
     max_active_node_calls: int = 16
 ```
 
-If this is too large for the first async parallel pass, leave global node-call budget as follow-up and rely on foreach `max_active`.
+If this is too large for the first concurrent pass, leave global node-call budget as follow-up and rely on foreach `max_active`.
 
 - [ ] **Step 8: Verify phase**
 
 Run:
 
 ```bash
-uv run pytest tests/core/test_parallel_foreach.py tests/authoring/test_demo_workflow.py -q
+uv run pytest tests/core/test_concurrent_foreach.py tests/authoring/test_demo_workflow.py -q
 uv run pytest -q
 uvx ruff check src tests
 uv run basedpyright --level error
@@ -554,12 +554,12 @@ Ship these as separate commits/PRs:
 1. Phase 1: policy shape and validation
 2. Phase 2: patch extraction with no behavior change
 3. Phase 3: barrier metadata with serial behavior unchanged
-4. Phase 4: async parallel execution
+4. Phase 4: concurrent execution
 
-Do not start Phase 4 until Phase 2 and Phase 3 are stable. Parallel foreach depends on patch extraction and resumable barrier state.
+Do not start Phase 4 until Phase 2 and Phase 3 are stable. Concurrent foreach depends on patch extraction and resumable barrier state.
 
 ## Self-Review
 
 - Spec coverage: ADR 0002 decisions are represented across the four phases.
 - Intentional gaps: explicit Fork/Gather, lineage-token graph nodes, OpenTelemetry, platform source/tool caps, and full run persistence are not included.
-- Risk control: phases 1-3 preserve serial behavior and keep `mode="parallel"` unsupported until phase 4.
+- Risk control: phases 1-3 preserve serial behavior and keep `mode="concurrent"` unsupported until phase 4.

docs/superpowers/plans/2026-05-22-scheduler-foundation.md

@@ -2,7 +2,7 @@
 
 > **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
 
-**Goal:** Add the internal scheduler foundation needed for future parallel foreach and native subgraphs while preserving current serial workflow behavior.
+**Goal:** Add the internal scheduler foundation needed for future concurrent foreach and native subgraphs while preserving current serial workflow behavior.
 
 **Architecture:** `RunState.frames` remains the source of frame lifecycle state, and a new serialized `ready_frame_ids` queue defines deterministic scheduling order. A new internal `wf_core.runtime.scheduler` module owns frame creation, enqueue/select, block/wake, and no-ready-frame resolution. Existing sync and async engines both use scheduler selection before `prepare_step`; node execution stays split between sync and async paths.
 
@@ -734,7 +734,7 @@ Expected: ruff passes and basedpyright reports 0 errors.
 Check:
 
 ```bash
-git diff -- CONTEXT.md docs/adr/0001-scheduler-foundation-before-parallel-foreach.md docs/current_roadmap.md
+git diff -- CONTEXT.md docs/adr/0001-scheduler-foundation-before-concurrent-foreach.md docs/current_roadmap.md
 ```
 
 Expected: docs remain aligned with implemented first-pass behavior.
@@ -744,5 +744,5 @@ Expected: docs remain aligned with implemented first-pass behavior.
 ## Self-Review
 
 - Spec coverage: ADR decisions are covered by tasks for `BLOCKED`, ready queue, scheduler helpers, sync/async engine migration, serial foreach block/wake, interrupt resume priority, and verification.
-- Intentional gaps: no `foreach(mode="parallel")`, no `ParallelForeachPolicy`, no lineage patches, no BarrierNode, and no public scheduler exports.
+- Intentional gaps: no `foreach(mode="concurrent")`, no `ForeachConcurrentPolicy`, no lineage patches, no BarrierNode, and no public scheduler exports.
 - Type consistency: helper names are stable across tasks: `add_frame`, `enqueue_frame`, `select_next_frame`, `mark_frame_pending`, `block_frame_on_children`, `wake_frame`, `wake_parent_if_children_complete`, and `resolve_no_ready_frames`.

docs/wf_core_architecture.md

@@ -77,8 +77,8 @@ enqueues the child. When the child reaches `END`, `wake_parent_if_children_compl
 wakes the blocked parent so it can create the next iteration or emit `done`.
 
 This preserves current serial behavior while making the hidden parent/child
-relationship explicit. `foreach(mode="parallel")` is still unsupported because
+relationship explicit. `foreach(mode="concurrent")` is still unsupported because
-parallel execution needs policy, barrier, lineage, and state-patch semantics
+concurrent execution needs policy, barrier, lineage, and state-patch semantics
 that are not implemented yet.
 
 ## Validation Flow
@@ -118,7 +118,7 @@ limits and intended adapter seam.
   state patch commits. The remaining mapping design notes for future reducer
   metadata are documented in
   [`core_state_mapping_and_merge.md`](core_state_mapping_and_merge.md).
-- Foreach is still serial-only. The scheduler foundation exists, but parallel
+- Foreach is still serial-only. The scheduler foundation exists, but concurrent
   foreach still needs explicit policy, implicit barrier state, lineage-aware
   patch commits, and quiescent interrupt handling. `ForeachNode.over` is typed
   as a `GraphSourcePath`, but execution is still serial.
@@ -137,9 +137,11 @@ limits and intended adapter seam.
   upgrade: nested run state, child-frame trace preservation, interrupt bubbling
   with path metadata, and resume back into the child workflow.
 - Frames are no longer only a serial execution stack: the runtime has a ready
-  queue and `BLOCKED` frame state. Async parallel foreach and native subgraphs
+  queue and `BLOCKED` frame state. Concurrent foreach and native subgraphs still
-  still need more work: lineage isolation, barrier merge semantics, pending
+  need more work: lineage isolation, barrier merge semantics, pending child
-  child results, and explicit child workflow/deployment identity.
+  results, and explicit child workflow/deployment identity. Async runtime can
+  later add simultaneous async node handler execution, but the workflow mode is
+  still concurrent foreach.
 - Runtime errors are still ordinary exceptions plus failed run status. A richer
   error payload can be added later, but should be designed as part of trace/run
   state rather than scattered exceptions.