zenprocess · zenprocess · Apr 8, 2026 · Apr 8, 2026
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -19,7 +19,6 @@ jobs:
             --exclude-loopback
             --max-redirects 5
             --accept '200..=204,206,301..=308,403,429'
-            --exclude 'github.com/zenprocess/switchyard'
             './**/*.md'
           fail: true
 

diff --git a/README.md b/README.md
@@ -29,7 +29,7 @@ The real bottleneck is **orchestration**: dispatching dozens of agents in parall
 | **Task** | Structured I/O — typed fields replace free-form prose. ~200 tokens vs ~2000. | **CACP** | [zenprocess/cacp](https://github.com/zenprocess/cacp) |
 | **Context** | 5-tier submodular file assignment, task-aware retrieval, cache-aware optimization. | **Sieeve** | _open-source release pending_ |
 | **Context (rules)** | Compact, tabular, LLM-native rule definition language. The format Sieeve compiles into. | **Axiom** | [zenprocess/axiom](https://github.com/zenprocess/axiom) |
-| **Dispatch** | Tracker-agnostic, git-native multi-agent dispatcher. Pre-dispatch doctor, response gate, AC compliance, regression detection. | **Switchyard** _(in development)_ | _personal R&D · open-source release pending_ |
+| **Dispatch** | Tracker-agnostic, git-native multi-agent dispatcher. Pre-dispatch doctor, response gate, AC compliance, regression detection. | _(internal — release pending)_ | _personal R&D_ |
 | **Verify** | 4-dimensional benchmark harness — multi-turn, multi-agent, parallel dispatch with tool calling. | **PawBench** | [zenprocess/pawbench](https://github.com/zenprocess/pawbench) |
 | **Verify (serving)** | Model registry for optimized LLM serving configurations. The "spec sheet" for self-hosted vLLM. | **ServingCard** | [zenprocess/servingcard](https://github.com/zenprocess/servingcard) |
 | **Learn** | Mines Claude Code session history for friction, patterns, gaps. Evolves your skills via autonomous experiment loops. | **Afterburn** | [zenprocess/afterburn](https://github.com/zenprocess/afterburn) |
@@ -40,7 +40,7 @@ The real bottleneck is **orchestration**: dispatching dozens of agents in parall
 
 ![Dispatch pipeline](diagrams/dispatch-flow.png)
 
-A single Switchyard dispatch end-to-end:
+A single dispatch, end-to-end:
 
 1. **Task** comes in (GitHub, Linear, GitLab — tracker-agnostic), wrapped in **CACP**
 2. **Sieeve** compiles task-aware context (Axiom rules + 5-tier file assignment)
@@ -85,11 +85,11 @@ Most AI coding work today is single-shot, manual, and unmeasured. Token costs ar
 
 standra.ai stands on the shoulders of researchers and builders who shipped the foundational ideas:
 
-- **[GEPA](https://gepa-ai.github.io/gepa/blog/2026/02/18/introducing-optimize-anything/)** — Goal/Effort/Plan/Assumptions reflective prompt optimization. Switchyard scores every dispatch on these 4 dimensions and feeds the verdict back into routing. *Agrawal et al., 2026*
+- **[GEPA](https://gepa-ai.github.io/gepa/blog/2026/02/18/introducing-optimize-anything/)** — Goal/Effort/Plan/Assumptions reflective prompt optimization. The dispatcher scores every dispatch on these 4 dimensions and feeds the verdict back into routing. *Agrawal et al., 2026*
 - **[RLM (Recursive Language Models)](https://arxiv.org/abs/2512.24601)** — *Zhang, Kraska & Khattab*. The REPL-driven recursive analysis pattern that lets Afterburn process 90MB+ session files no context window can hold.
 - **[CCAR (Claude Code AutoResearch)](https://github.com/mitkox/ccar)** — *Mitko Vasilev*. The autonomous experiment loop Afterburn uses to evolve skills (mutate → benchmark → keep/discard → commit).
 - **[Andrej Karpathy](https://karpathy.ai/)** — *Software 2.0*, *LLM OS*, and the public thinking on agentic systems that frames why orchestration is the real engineering problem, not the model.
-- **[addyosmani/agent-skills](https://github.com/addyosmani/agent-skills)** — *Addy Osmani*. The Save Point Pattern, anti-rationalization tables, and skills-as-gated-workflows pattern that shape how Switchyard structures dispatch prompts.
+- **[addyosmani/agent-skills](https://github.com/addyosmani/agent-skills)** — *Addy Osmani*. The Save Point Pattern, anti-rationalization tables, and skills-as-gated-workflows pattern that shape how the dispatcher structures dispatch prompts.
 - **[DSPy](https://dspy.ai/)** — *Stanford NLP*. The original GEPA optimizer and the systematic approach to LM programming that influenced CACP's structured I/O design.
 - **[LLMLingua](https://github.com/microsoft/LLMLingua)** — *Microsoft Research*. Prompt compression research underpinning Sieeve's tiered context strategy.
 

diff --git a/docs/design-system.md b/docs/design-system.md
@@ -15,7 +15,7 @@ The standra.ai parent brand uses a **blue → purple → magenta** gradient. Eve
 | Project | Stop 1 | Stop 2 | Stop 3 | Stop 4 | Metaphor |
 |---|---|---|---|---|---|
 | **standra** (parent) | `#0066ff` | `#5544ff` | `#aa44ff` | `#ff44dd` | platform / hub |
-| **Switchyard** | `#0066ff` | `#3388ff` | `#55aaff` | `#88ccff` | rail / signal |
+| **Dispatch** | `#0066ff` | `#3388ff` | `#55aaff` | `#88ccff` | rail / signal |
 | **Sieeve** | `#8833ff` | `#9955ff` | `#aa77ff` | `#cc99ff` | filter / prism |
 | **Axiom** | `#6644dd` | `#8866ee` | `#aa88ff` | `#ccaaff` | rule / law |
 | **CACP** | `#3344ff` | `#5566ff` | `#7788ff` | `#99aaff` | protocol / wire |

diff --git a/docs/diagram-spec.md b/docs/diagram-spec.md
@@ -49,7 +49,7 @@ Derived from the brand axis in [design-system.md](design-system.md). Light fill
 | Project | Fill | Stroke | Notes |
 |---|---|---|---|
 | **standra (parent hub)** | `#0066ff` | `#003388` | Saturated fill, white text inside, stroke width 4 |
-| **Switchyard** | `#cce4ff` | `#0066ff` | Stroke width 3 when shown as a hub |
+| **Dispatch** | `#cce4ff` | `#0066ff` | Stroke width 3 when shown as a hub |
 | **Sieeve** | `#ead4ff` | `#8833ff` | |
 | **Axiom** | `#e4d9ff` | `#6644dd` | |
 | **CACP** | `#cad0ff` | `#3344ff` | |
@@ -67,7 +67,7 @@ Use these inside intra-project diagrams (where boxes represent roles, not projec
 
 | Semantic role | Fill | Stroke | Use for |
 |---|---|---|---|
-| **Hub / primary** | `#3b82f6` | `#1e3a5f` | The thing the diagram is about. Switchyard in the ecosystem map. |
+| **Hub / primary** | `#3b82f6` | `#1e3a5f` | The thing the diagram is about. Dispatch in the ecosystem map. |
 | **Inputs / context** | `#ddd6fe` | `#6d28d9` | CACP, Sieeve, Axiom — anything feeding context in. |
 | **Outputs / success** | `#a7f3d0` | `#047857` | Diagnostics, telemetry, merged results. |
 | **Start / trigger** | `#fed7aa` | `#c2410c` | The initiating task or event. |
@@ -107,7 +107,7 @@ Use the visual pattern that mirrors the concept's behaviour. **Each major concep
 
 | Concept | Pattern | Example |
 |---|---|---|
-| Many things converge on one | **Hub-and-spoke** (radial) | Switchyard with CACP, Sieeve, Axiom feeding in |
+| Many things converge on one | **Hub-and-spoke** (radial) | Dispatch with CACP, Sieeve, Axiom feeding in |
 | Linear sequence of stages | **Horizontal pipeline** | Task → Context → Dispatch → Verify → Merge |
 | Loop / continuous improvement | **Cycle** with return arrow | Telemetry → Afterburn → Sieeve rules |
 | Decision branching | **Diamond** with labelled outgoing edges | Doctor with pass/fail |

diff --git a/docs/og-spec.md b/docs/og-spec.md
@@ -156,7 +156,7 @@ Everything else (layout, typography, grid, badge, effects) stays identical.
 | Project | Stop 1 | Stop 2 | Stop 3 | Stop 4 | Metaphor |
 |---------|--------|--------|--------|--------|----------|
 | Afterburn | #ff4500 | #ff6b35 | #ffa500 | #ffd700 | Fire |
-| Switchyard | #0066ff | #3388ff | #55aaff | #88ccff | Rail/signal |
+| Dispatch | #0066ff | #3388ff | #55aaff | #88ccff | Rail/signal |
 | Sentinel | #00cc66 | #33dd88 | #66eeaa | #99ffcc | Shield/scan |
 | Sieeve | #8833ff | #9955ff | #aa77ff | #cc99ff | Filter/prism |
 | ServingCard | #ff3366 | #ff5588 | #ff77aa | #ff99cc | Card/benchmark |