diff --git a/.github/workflows/daily-fact.lock.yml b/.github/workflows/daily-fact.lock.yml
index eaa4424d84d..d2918239aaf 100644
--- a/.github/workflows/daily-fact.lock.yml
+++ b/.github/workflows/daily-fact.lock.yml
@@ -1,4 +1,4 @@
-# gh-aw-metadata: {"schema_version":"v3","frontmatter_hash":"af7c61e76abdaae9d6d291e56f257ddcba114e4dc0595519936f5beadafcffec","strict":true,"agent_id":"codex","agent_model":"gpt-5.4-mini"}
+# gh-aw-metadata: {"schema_version":"v3","frontmatter_hash":"36d37a405b99aa890e631983125c296a231d492607435c20e8632d45621f8ecd","strict":true,"agent_id":"codex","agent_model":"gpt-5.4-mini"}
 # gh-aw-manifest: {"version":1,"secrets":["CODEX_API_KEY","GH_AW_GITHUB_MCP_SERVER_TOKEN","GH_AW_GITHUB_TOKEN","GH_AW_OTEL_ENDPOINT","GH_AW_OTEL_HEADERS","GITHUB_TOKEN","OPENAI_API_KEY"],"actions":[{"repo":"actions/cache/restore","sha":"27d5ce7f107fe9357f9df03efb73ab90386fccae","version":"v5.0.5"},{"repo":"actions/cache/save","sha":"27d5ce7f107fe9357f9df03efb73ab90386fccae","version":"v5.0.5"},{"repo":"actions/checkout","sha":"de0fac2e4500dabe0009e67214ff5f5447ce83dd","version":"v6.0.2"},{"repo":"actions/download-artifact","sha":"3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c","version":"v8.0.1"},{"repo":"actions/github-script","sha":"3a2844b7e9c422d3c10d287c895573f7108da1b3","version":"v9.0.0"},{"repo":"actions/setup-node","sha":"48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e","version":"v6.4.0"},{"repo":"actions/setup-python","sha":"a309ff8b426b58ec0e2a45f0f869d46889d02405","version":"v6.2.0"},{"repo":"actions/upload-artifact","sha":"043fb46d1a93c77aae656e7c1c64a875d1fc6a0a","version":"v7.0.1"}],"containers":[{"image":"ghcr.io/github/gh-aw-firewall/agent:0.25.42"},{"image":"ghcr.io/github/gh-aw-firewall/api-proxy:0.25.42"},{"image":"ghcr.io/github/gh-aw-firewall/cli-proxy:0.25.42"},{"image":"ghcr.io/github/gh-aw-firewall/squid:0.25.42"},{"image":"ghcr.io/github/gh-aw-mcpg:v0.3.6","digest":"sha256:2bb8eef86006a4c5963c55616a9c51c32f27bfdecb023b8aa6f91f6718d9171c","pinned_image":"ghcr.io/github/gh-aw-mcpg:v0.3.6@sha256:2bb8eef86006a4c5963c55616a9c51c32f27bfdecb023b8aa6f91f6718d9171c"},{"image":"ghcr.io/github/github-mcp-server:v1.0.3","digest":"sha256:2ac27ef03461ef2b877031b838a7d1fd7f12b12d4ace7796d8cad91446d55959","pinned_image":"ghcr.io/github/github-mcp-server:v1.0.3@sha256:2ac27ef03461ef2b877031b838a7d1fd7f12b12d4ace7796d8cad91446d55959"},{"image":"node:lts-alpine","digest":"sha256:d1b3b4da11eefd5941e7f0b9cf17783fc99d9c6fc34884a665f40a06dbdfc94f","pinned_image":"node:lts-alpine@sha256:d1b3b4da11eefd5941e7f0b9cf17783fc99d9c6fc34884a665f40a06dbdfc94f"}]}
 #    ___                   _   _      
 #   / _ \                 | | (_)     
@@ -97,6 +97,7 @@ jobs:
       comment_id: ""
       comment_repo: ""
       engine_id: ${{ steps.generate_aw_info.outputs.engine_id }}
+      experiments: ${{ steps.pick-experiment.outputs.experiments }}
       lockdown_check_failed: ${{ steps.generate_aw_info.outputs.lockdown_check_failed == 'true' }}
       model: ${{ steps.generate_aw_info.outputs.model }}
       secret_verification_result: ${{ steps.validate-secret.outputs.verification_result }}
@@ -189,10 +190,46 @@ jobs:
             setupGlobals(core, github, context, exec, io, getOctokit);
             const { main } = require('${{ runner.temp }}/gh-aw/actions/check_workflow_timestamp_api.cjs');
             await main();
+      - name: Restore experiment state from git
+        id: restore-experiment-state
+        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
+        env:
+          GH_AW_EXPERIMENT_STATE_FILE: /tmp/gh-aw/experiments/state.json
+          GH_AW_EXPERIMENT_STATE_DIR: /tmp/gh-aw/experiments
+          GH_AW_EXPERIMENT_BRANCH: experiments/dailyfact
+        with:
+          script: |
+            const { setupGlobals } = require('${{ runner.temp }}/gh-aw/actions/setup_globals.cjs');
+            setupGlobals(core, github, context, exec, io, getOctokit);
+            const { main } = require('${{ runner.temp }}/gh-aw/actions/load_experiment_state_from_repo.cjs');
+            await main();
+      - name: Pick experiment variants
+        id: pick-experiment
+        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
+        env:
+          GH_AW_EXPERIMENT_SPEC: '{"reasoning_depth":{"variants":["single_pass","multi_candidate"],"description":"Tests whether deliberating over multiple candidate facts before writing improves verse novelty and engagement.","hypothesis":"H0: no change in discussion engagement rate. H1: multi_candidate produces more novel verses with higher reaction counts (expected +20% reactions).","metric":"discussion_reaction_count","secondary_metrics":["output_length_chars","run_duration_ms"],"guardrail_metrics":[{"name":"empty_output_rate","threshold":"\u003c0.05"},{"name":"run_success_rate","threshold":"\u003e=0.95"}],"min_samples":30,"weight":[50,50],"issue":31324,"start_date":"2026-05-11"}}'
+          GH_AW_EXPERIMENT_STATE_FILE: /tmp/gh-aw/experiments/state.json
+          GH_AW_EXPERIMENT_STATE_DIR: /tmp/gh-aw/experiments
+        with:
+          script: |
+            const { setupGlobals } = require('${{ runner.temp }}/gh-aw/actions/setup_globals.cjs');
+            setupGlobals(core, github, context, exec, io, getOctokit);
+            const { main } = require('${{ runner.temp }}/gh-aw/actions/pick_experiment.cjs');
+            await main();
+      - name: Upload experiment artifact
+        if: always()
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: dailyfact-experiment
+          path: /tmp/gh-aw/experiments
+          if-no-files-found: ignore
+          retention-days: 30
       - name: Create prompt with built-in context
         env:
           GH_AW_PROMPT: /tmp/gh-aw/aw-prompts/prompt.txt
           GH_AW_SAFE_OUTPUTS: ${{ runner.temp }}/gh-aw/safeoutputs/outputs.jsonl
+          GH_AW_EXPERIMENTS_REASONING_DEPTH: ${{ steps.pick-experiment.outputs.reasoning_depth }}
+          GH_AW_EXPR_DD64DF46: ${{ steps.pick-experiment.outputs.reasoning_depth == "multi_candidate" }}
           GH_AW_GITHUB_ACTOR: ${{ github.actor }}
           GH_AW_GITHUB_EVENT_COMMENT_ID: ${{ github.event.comment.id }}
           GH_AW_GITHUB_EVENT_DISCUSSION_NUMBER: ${{ github.event.discussion.number }}
@@ -205,21 +242,21 @@ jobs:
         run: |
           bash "${RUNNER_TEMP}/gh-aw/actions/create_prompt_first.sh"
           {
-          cat << 'GH_AW_PROMPT_291a7ecc0036fade_EOF'
+          cat << 'GH_AW_PROMPT_6f6003e03d2c7d0e_EOF'
           <system>
-          GH_AW_PROMPT_291a7ecc0036fade_EOF
+          GH_AW_PROMPT_6f6003e03d2c7d0e_EOF
           cat "${RUNNER_TEMP}/gh-aw/prompts/xpia.md"
           cat "${RUNNER_TEMP}/gh-aw/prompts/temp_folder_prompt.md"
           cat "${RUNNER_TEMP}/gh-aw/prompts/markdown.md"
           cat "${RUNNER_TEMP}/gh-aw/prompts/cache_memory_prompt.md"
           cat "${RUNNER_TEMP}/gh-aw/prompts/safe_outputs_prompt.md"
-          cat << 'GH_AW_PROMPT_291a7ecc0036fade_EOF'
+          cat << 'GH_AW_PROMPT_6f6003e03d2c7d0e_EOF'
           <safe-output-tools>
           Tools: add_comment, missing_tool, missing_data, noop
           </safe-output-tools>
-          GH_AW_PROMPT_291a7ecc0036fade_EOF
+          GH_AW_PROMPT_6f6003e03d2c7d0e_EOF
           cat "${RUNNER_TEMP}/gh-aw/prompts/mcp_cli_tools_prompt.md"
-          cat << 'GH_AW_PROMPT_291a7ecc0036fade_EOF'
+          cat << 'GH_AW_PROMPT_6f6003e03d2c7d0e_EOF'
           <github-context>
           The following GitHub context information is available for this workflow:
           {{#if __GH_AW_GITHUB_ACTOR__ }}
@@ -248,9 +285,9 @@ jobs:
           {{/if}}
           </github-context>
           
-          GH_AW_PROMPT_291a7ecc0036fade_EOF
+          GH_AW_PROMPT_6f6003e03d2c7d0e_EOF
           cat "${RUNNER_TEMP}/gh-aw/prompts/cli_proxy_with_safeoutputs_prompt.md"
-          cat << 'GH_AW_PROMPT_291a7ecc0036fade_EOF'
+          cat << 'GH_AW_PROMPT_6f6003e03d2c7d0e_EOF'
           </system>
           
           
@@ -297,7 +334,11 @@ jobs:
           ## Guidelines
           
           - **Check memory first**: Skip any PR, issue, or release that already appears in the palace results from Step 0
+          {{#if __GH_AW_EXPR_DD64DF46__ }}
+          - **Multi-candidate deliberation**: Before writing, identify exactly **3 distinct candidate facts** (one PR, one issue or release, one contributor or pattern). For each candidate write one sentence on why it is novel today. Then score each candidate 1–5 on: (a) novelty vs palace memory, (b) intrinsic poetic potential. Select the highest-scoring candidate and write the verse for that one only.
+          {{else}}
           - **Favor recent updates** but include variety - pick something interesting, not just the most recent
+          {{/if}}
           - **Be specific**: Include PR numbers, issue references, or release tags when relevant
           - **Keep it short**: One or two poetic sentences for the main fact, optionally with a brief context
           - **Be poetic**: Use lyrical, whimsical language that celebrates the beauty of code and collaboration
@@ -349,13 +390,15 @@ jobs:
           
           {{#runtime-import shared/noop-reminder.md}}
           
-          GH_AW_PROMPT_291a7ecc0036fade_EOF
+          GH_AW_PROMPT_6f6003e03d2c7d0e_EOF
           } > "$GH_AW_PROMPT"
       - name: Interpolate variables and render templates
         uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
         env:
           GH_AW_PROMPT: /tmp/gh-aw/aw-prompts/prompt.txt
           GH_AW_ENGINE_ID: "codex"
+          GH_AW_EXPERIMENTS_REASONING_DEPTH: ${{ steps.pick-experiment.outputs.reasoning_depth }}
+          GH_AW_EXPR_DD64DF46: ${{ steps.pick-experiment.outputs.reasoning_depth == "multi_candidate" }}
           GH_AW_GITHUB_REPOSITORY: ${{ github.repository }}
         with:
           script: |
@@ -370,6 +413,8 @@ jobs:
           GH_AW_ALLOWED_EXTENSIONS: ''
           GH_AW_CACHE_DESCRIPTION: ''
           GH_AW_CACHE_DIR: '/tmp/gh-aw/cache-memory/'
+          GH_AW_EXPERIMENTS_REASONING_DEPTH: ${{ steps.pick-experiment.outputs.reasoning_depth }}
+          GH_AW_EXPR_DD64DF46: ${{ steps.pick-experiment.outputs.reasoning_depth == "multi_candidate" }}
           GH_AW_GITHUB_ACTOR: ${{ github.actor }}
           GH_AW_GITHUB_EVENT_COMMENT_ID: ${{ github.event.comment.id }}
           GH_AW_GITHUB_EVENT_DISCUSSION_NUMBER: ${{ github.event.discussion.number }}
@@ -393,6 +438,8 @@ jobs:
                 GH_AW_ALLOWED_EXTENSIONS: process.env.GH_AW_ALLOWED_EXTENSIONS,
                 GH_AW_CACHE_DESCRIPTION: process.env.GH_AW_CACHE_DESCRIPTION,
                 GH_AW_CACHE_DIR: process.env.GH_AW_CACHE_DIR,
+                GH_AW_EXPERIMENTS_REASONING_DEPTH: process.env.GH_AW_EXPERIMENTS_REASONING_DEPTH,
+                GH_AW_EXPR_DD64DF46: process.env.GH_AW_EXPR_DD64DF46,
                 GH_AW_GITHUB_ACTOR: process.env.GH_AW_GITHUB_ACTOR,
                 GH_AW_GITHUB_EVENT_COMMENT_ID: process.env.GH_AW_GITHUB_EVENT_COMMENT_ID,
                 GH_AW_GITHUB_EVENT_DISCUSSION_NUMBER: process.env.GH_AW_GITHUB_EVENT_DISCUSSION_NUMBER,
@@ -598,9 +645,9 @@ jobs:
           mkdir -p "${RUNNER_TEMP}/gh-aw/safeoutputs"
           mkdir -p /tmp/gh-aw/safeoutputs
           mkdir -p /tmp/gh-aw/mcp-logs/safeoutputs
-          cat > "${RUNNER_TEMP}/gh-aw/safeoutputs/config.json" << 'GH_AW_SAFE_OUTPUTS_CONFIG_624158a7bed00a1d_EOF'
+          cat > "${RUNNER_TEMP}/gh-aw/safeoutputs/config.json" << 'GH_AW_SAFE_OUTPUTS_CONFIG_d1209a06df4079df_EOF'
           {"add_comment":{"max":1,"target":"4750"},"create_report_incomplete_issue":{},"missing_data":{},"missing_tool":{},"noop":{"max":1,"report-as-issue":"true"},"report_incomplete":{}}
-          GH_AW_SAFE_OUTPUTS_CONFIG_624158a7bed00a1d_EOF
+          GH_AW_SAFE_OUTPUTS_CONFIG_d1209a06df4079df_EOF
       - name: Generate Safe Outputs Tools
         env:
           GH_AW_TOOLS_META_JSON: |
@@ -787,7 +834,7 @@ jobs:
           DOCKER_SOCK_GID=$(stat -c '%g' /var/run/docker.sock 2>/dev/null || echo '0')
           export MCP_GATEWAY_DOCKER_COMMAND='docker run -i --rm --network host --add-host host.docker.internal:127.0.0.1 --user '"${MCP_GATEWAY_UID}"':'"${MCP_GATEWAY_GID}"' --group-add '"${DOCKER_SOCK_GID}"' -v /var/run/docker.sock:/var/run/docker.sock -e MCP_GATEWAY_PORT -e MCP_GATEWAY_DOMAIN -e MCP_GATEWAY_API_KEY -e MCP_GATEWAY_PAYLOAD_DIR -e MCP_GATEWAY_PAYLOAD_SIZE_THRESHOLD -e DEBUG -e MCP_GATEWAY_LOG_DIR -e GH_AW_MCP_LOG_DIR -e GH_AW_SAFE_OUTPUTS -e GH_AW_SAFE_OUTPUTS_CONFIG_PATH -e GH_AW_SAFE_OUTPUTS_TOOLS_PATH -e GH_AW_ASSETS_BRANCH -e GH_AW_ASSETS_MAX_SIZE_KB -e GH_AW_ASSETS_ALLOWED_EXTS -e DEFAULT_BRANCH -e GITHUB_MCP_SERVER_TOKEN -e GITHUB_MCP_GUARD_MIN_INTEGRITY -e GITHUB_MCP_GUARD_REPOS -e GITHUB_REPOSITORY -e GITHUB_SERVER_URL -e GITHUB_SHA -e GITHUB_WORKSPACE -e GITHUB_TOKEN -e GITHUB_RUN_ID -e GITHUB_RUN_NUMBER -e GITHUB_RUN_ATTEMPT -e GITHUB_JOB -e GITHUB_ACTION -e GITHUB_EVENT_NAME -e GITHUB_EVENT_PATH -e GITHUB_ACTOR -e GITHUB_ACTOR_ID -e GITHUB_TRIGGERING_ACTOR -e GITHUB_WORKFLOW -e GITHUB_WORKFLOW_REF -e GITHUB_WORKFLOW_SHA -e GITHUB_REF -e GITHUB_REF_NAME -e GITHUB_REF_TYPE -e GITHUB_HEAD_REF -e GITHUB_BASE_REF -e GH_AW_SAFE_OUTPUTS_PORT -e GH_AW_SAFE_OUTPUTS_API_KEY -e GITHUB_AW_OTEL_TRACE_ID -e GITHUB_AW_OTEL_PARENT_SPAN_ID -e CODEX_HOME -v /tmp/gh-aw/mcp-payloads:/tmp/gh-aw/mcp-payloads:rw -v /opt:/opt:ro -v /tmp:/tmp:rw -v '"${GITHUB_WORKSPACE}"':'"${GITHUB_WORKSPACE}"':rw ghcr.io/github/gh-aw-mcpg:v0.3.6'
           
-          cat > "${RUNNER_TEMP}/gh-aw/mcp-config/config.toml" << GH_AW_MCP_CONFIG_df65c8ac8eef6bf5_EOF
+          cat > "${RUNNER_TEMP}/gh-aw/mcp-config/config.toml" << GH_AW_MCP_CONFIG_ece71de4bad38321_EOF
           [history]
           persistence = "none"
           
@@ -814,11 +861,11 @@ jobs:
           
           [mcp_servers.safeoutputs."guard-policies".write-sink]
           accept = ["*"]
-          GH_AW_MCP_CONFIG_df65c8ac8eef6bf5_EOF
+          GH_AW_MCP_CONFIG_ece71de4bad38321_EOF
           
           # Generate JSON config for MCP gateway
           GH_AW_NODE=$(which node 2>/dev/null || command -v node 2>/dev/null || echo node)
-          cat << GH_AW_MCP_CONFIG_df65c8ac8eef6bf5_EOF | "$GH_AW_NODE" "${RUNNER_TEMP}/gh-aw/actions/start_mcp_gateway.cjs"
+          cat << GH_AW_MCP_CONFIG_ece71de4bad38321_EOF | "$GH_AW_NODE" "${RUNNER_TEMP}/gh-aw/actions/start_mcp_gateway.cjs"
           {
             "mcpServers": {
               "mempalace": {
@@ -881,11 +928,11 @@ jobs:
               }
             }
           }
-          GH_AW_MCP_CONFIG_df65c8ac8eef6bf5_EOF
+          GH_AW_MCP_CONFIG_ece71de4bad38321_EOF
           
           # Sync converter output to writable CODEX_HOME for Codex
           mkdir -p /tmp/gh-aw/mcp-config
-          cat > "/tmp/gh-aw/mcp-config/config.toml" << GH_AW_CODEX_SHELL_POLICY_01ccd543b3815814_EOF
+          cat > "/tmp/gh-aw/mcp-config/config.toml" << GH_AW_CODEX_SHELL_POLICY_50e3008d43a21862_EOF
 
           model_provider = "openai-proxy"
 
@@ -897,7 +944,7 @@ jobs:
           [shell_environment_policy]
           inherit = "core"
           include_only = ["CODEX_API_KEY", "GH_AW_ASSETS_ALLOWED_EXTS", "GH_AW_ASSETS_BRANCH", "GH_AW_ASSETS_MAX_SIZE_KB", "GH_AW_SAFE_OUTPUTS", "GITHUB_REPOSITORY", "GITHUB_SERVER_URL", "HOME", "OPENAI_API_KEY", "PATH"]
-          GH_AW_CODEX_SHELL_POLICY_01ccd543b3815814_EOF
+          GH_AW_CODEX_SHELL_POLICY_50e3008d43a21862_EOF
           awk '
             BEGIN { skip_openai_proxy = 0 }
             /^[[:space:]]*model_provider[[:space:]]*=/ { next }
@@ -1148,6 +1195,7 @@ jobs:
       - activation
       - agent
       - detection
+      - push_experiments_state
       - safe_outputs
       - update_cache_memory
     if: >
@@ -1349,6 +1397,12 @@ jobs:
           mkdir -p /tmp/gh-aw/
           find "/tmp/gh-aw/" -type f -print
           echo "GH_AW_AGENT_OUTPUT=/tmp/gh-aw/agent_output.json" >> "$GITHUB_OUTPUT"
+      - name: Download experiment artifact
+        continue-on-error: true
+        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
+        with:
+          name: dailyfact-experiment
+          path: /tmp/gh-aw/experiments/
       - name: Checkout repository for patch context
         if: needs.agent.outputs.has_patch == 'true'
         uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
@@ -1448,18 +1502,18 @@ jobs:
           DOCKER_SOCK_GID=$(stat -c '%g' /var/run/docker.sock 2>/dev/null || echo '0')
           export MCP_GATEWAY_DOCKER_COMMAND='docker run -i --rm --network host --add-host host.docker.internal:127.0.0.1 --user '"${MCP_GATEWAY_UID}"':'"${MCP_GATEWAY_GID}"' --group-add '"${DOCKER_SOCK_GID}"' -v /var/run/docker.sock:/var/run/docker.sock -e MCP_GATEWAY_PORT -e MCP_GATEWAY_DOMAIN -e MCP_GATEWAY_API_KEY -e MCP_GATEWAY_PAYLOAD_DIR -e MCP_GATEWAY_PAYLOAD_SIZE_THRESHOLD -e DEBUG -e MCP_GATEWAY_LOG_DIR -e GH_AW_MCP_LOG_DIR -e GH_AW_SAFE_OUTPUTS -e GH_AW_SAFE_OUTPUTS_CONFIG_PATH -e GH_AW_SAFE_OUTPUTS_TOOLS_PATH -e GH_AW_ASSETS_BRANCH -e GH_AW_ASSETS_MAX_SIZE_KB -e GH_AW_ASSETS_ALLOWED_EXTS -e DEFAULT_BRANCH -e GITHUB_MCP_SERVER_TOKEN -e GITHUB_MCP_GUARD_MIN_INTEGRITY -e GITHUB_MCP_GUARD_REPOS -e GITHUB_REPOSITORY -e GITHUB_SERVER_URL -e GITHUB_SHA -e GITHUB_WORKSPACE -e GITHUB_TOKEN -e GITHUB_RUN_ID -e GITHUB_RUN_NUMBER -e GITHUB_RUN_ATTEMPT -e GITHUB_JOB -e GITHUB_ACTION -e GITHUB_EVENT_NAME -e GITHUB_EVENT_PATH -e GITHUB_ACTOR -e GITHUB_ACTOR_ID -e GITHUB_TRIGGERING_ACTOR -e GITHUB_WORKFLOW -e GITHUB_WORKFLOW_REF -e GITHUB_WORKFLOW_SHA -e GITHUB_REF -e GITHUB_REF_NAME -e GITHUB_REF_TYPE -e GITHUB_HEAD_REF -e GITHUB_BASE_REF -e CODEX_HOME -v /tmp/gh-aw/mcp-payloads:/tmp/gh-aw/mcp-payloads:rw -v /opt:/opt:ro -v /tmp:/tmp:rw -v '"${GITHUB_WORKSPACE}"':'"${GITHUB_WORKSPACE}"':rw ghcr.io/github/gh-aw-mcpg:v0.3.6'
           
-          cat > "${RUNNER_TEMP}/gh-aw/mcp-config/config.toml" << GH_AW_MCP_CONFIG_0a43f5caf239327a_EOF
+          cat > "${RUNNER_TEMP}/gh-aw/mcp-config/config.toml" << GH_AW_MCP_CONFIG_d3684e3b46c26350_EOF
           [history]
           persistence = "none"
           
           [shell_environment_policy]
           inherit = "core"
           include_only = ["CODEX_API_KEY", "HOME", "OPENAI_API_KEY", "PATH"]
-          GH_AW_MCP_CONFIG_0a43f5caf239327a_EOF
+          GH_AW_MCP_CONFIG_d3684e3b46c26350_EOF
           
           # Generate JSON config for MCP gateway
           GH_AW_NODE=$(which node 2>/dev/null || command -v node 2>/dev/null || echo node)
-          cat << GH_AW_MCP_CONFIG_2074415d4870b663_EOF | "$GH_AW_NODE" "${RUNNER_TEMP}/gh-aw/actions/start_mcp_gateway.cjs"
+          cat << GH_AW_MCP_CONFIG_6e95c2a80ec5cd9d_EOF | "$GH_AW_NODE" "${RUNNER_TEMP}/gh-aw/actions/start_mcp_gateway.cjs"
           {
             "mcpServers": {
             },
@@ -1470,11 +1524,11 @@ jobs:
               "payloadDir": "${MCP_GATEWAY_PAYLOAD_DIR}"
             }
           }
-          GH_AW_MCP_CONFIG_2074415d4870b663_EOF
+          GH_AW_MCP_CONFIG_6e95c2a80ec5cd9d_EOF
           
           # Sync converter output to writable CODEX_HOME for Codex
           mkdir -p /tmp/gh-aw/mcp-config
-          cat > "/tmp/gh-aw/mcp-config/config.toml" << GH_AW_CODEX_SHELL_POLICY_3bc3ee739bd056b7_EOF
+          cat > "/tmp/gh-aw/mcp-config/config.toml" << GH_AW_CODEX_SHELL_POLICY_71d735b59c609993_EOF
           model_provider = "openai-proxy"
           [model_providers.openai-proxy]
           name = "OpenAI AWF proxy"
@@ -1484,7 +1538,7 @@ jobs:
           [shell_environment_policy]
           inherit = "core"
           include_only = ["CODEX_API_KEY", "HOME", "OPENAI_API_KEY", "PATH"]
-          GH_AW_CODEX_SHELL_POLICY_3bc3ee739bd056b7_EOF
+          GH_AW_CODEX_SHELL_POLICY_71d735b59c609993_EOF
           awk '
             BEGIN { skip_openai_proxy = 0 }
             /^[[:space:]]*model_provider[[:space:]]*=/ { next }
@@ -1564,6 +1618,82 @@ jobs:
               }
             }
 
+  push_experiments_state:
+    needs: activation
+    if: always() && (!cancelled()) && needs.activation.result == 'success'
+    runs-on: ubuntu-slim
+    permissions:
+      contents: write
+    steps:
+      - name: Checkout actions folder
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          repository: github/gh-aw
+          sparse-checkout: |
+            actions
+          persist-credentials: false
+      - name: Setup Scripts
+        id: setup
+        uses: ./actions/setup
+        with:
+          destination: ${{ runner.temp }}/gh-aw/actions
+          job-name: ${{ github.job }}
+          trace-id: ${{ needs.activation.outputs.setup-trace-id }}
+          parent-span-id: ${{ needs.activation.outputs.setup-parent-span-id || needs.activation.outputs.setup-span-id }}
+        env:
+          GH_AW_SETUP_WORKFLOW_NAME: "Daily Fact About gh-aw"
+          GH_AW_CURRENT_WORKFLOW_REF: ${{ github.repository }}/.github/workflows/daily-fact.lock.yml@${{ github.ref }}
+          GH_AW_INFO_VERSION: "0.129.0"
+      - name: Checkout repository
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          persist-credentials: false
+          sparse-checkout: .
+      - name: Configure Git credentials
+        env:
+          REPO_NAME: ${{ github.repository }}
+          SERVER_URL: ${{ github.server_url }}
+          GITHUB_TOKEN: ${{ github.token }}
+        run: |
+          git config --global user.email "github-actions[bot]@users.noreply.github.com"
+          git config --global user.name "github-actions[bot]"
+          git config --global am.keepcr true
+          # Re-authenticate git with GitHub token
+          SERVER_URL_STRIPPED="${SERVER_URL#https://}"
+          git remote set-url origin "https://x-access-token:${GITHUB_TOKEN}@${SERVER_URL_STRIPPED}/${REPO_NAME}.git"
+          echo "Git configured with standard GitHub Actions identity"
+      - name: Download experiment artifact
+        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
+        continue-on-error: true
+        with:
+          name: dailyfact-experiment
+          path: /tmp/gh-aw/experiments
+      - name: Push experiment state to git
+        id: push_experiments_state
+        if: always()
+        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
+        env:
+          GH_TOKEN: ${{ github.token }}
+          GITHUB_RUN_ID: ${{ github.run_id }}
+          GITHUB_SERVER_URL: ${{ github.server_url }}
+          GH_AW_EXPERIMENT_STATE_DIR: /tmp/gh-aw/experiments
+          GH_AW_EXPERIMENT_BRANCH: experiments/dailyfact
+        with:
+          script: |
+            const { setupGlobals } = require('${{ runner.temp }}/gh-aw/actions/setup_globals.cjs');
+            setupGlobals(core, github, context, exec, io, getOctokit);
+            const { main } = require('${{ runner.temp }}/gh-aw/actions/push_experiment_state.cjs');
+            await main();
+      - name: Restore actions folder
+        if: always()
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          repository: github/gh-aw
+          sparse-checkout: |
+            actions/setup
+          sparse-checkout-cone-mode: true
+          persist-credentials: false
+
   safe_outputs:
     needs:
       - activation
diff --git a/.github/workflows/daily-fact.md b/.github/workflows/daily-fact.md
index 434d3e57b89..7574d87991c 100644
--- a/.github/workflows/daily-fact.md
+++ b/.github/workflows/daily-fact.md
@@ -16,6 +16,22 @@ engine:
   model: gpt-5.4-mini
   bare: true
 strict: true
+experiments:
+  reasoning_depth:
+    variants: [single_pass, multi_candidate]
+    description: "Tests whether deliberating over multiple candidate facts before writing improves verse novelty and engagement."
+    hypothesis: "H0: no change in discussion engagement rate. H1: multi_candidate produces more novel verses with higher reaction counts (expected +20% reactions)."
+    metric: discussion_reaction_count
+    secondary_metrics: [output_length_chars, run_duration_ms]
+    guardrail_metrics:
+      - name: empty_output_rate
+        threshold: "<0.05"
+      - name: run_success_rate
+        threshold: ">=0.95"
+    min_samples: 30
+    weight: [50, 50]
+    start_date: "2026-05-11"
+    issue: 31324
 timeout-minutes: 15
 runs-on: aw-gpu-runner-T4
 runtimes:
@@ -83,7 +99,11 @@ Mine recent activity from the repository to find interesting facts. Focus on:
 ## Guidelines
 
 - **Check memory first**: Skip any PR, issue, or release that already appears in the palace results from Step 0
+{{#if experiments.reasoning_depth == "multi_candidate"}}
+- **Multi-candidate deliberation**: Before writing, identify exactly **3 distinct candidate facts** (one PR, one issue or release, one contributor or pattern). For each candidate write one sentence on why it is novel today. Then score each candidate 1–5 on: (a) novelty vs palace memory, (b) intrinsic poetic potential. Select the highest-scoring candidate and write the verse for that one only.
+{{else}}
 - **Favor recent updates** but include variety - pick something interesting, not just the most recent
+{{/if}}
 - **Be specific**: Include PR numbers, issue references, or release tags when relevant
 - **Keep it short**: One or two poetic sentences for the main fact, optionally with a brief context
 - **Be poetic**: Use lyrical, whimsical language that celebrates the beauty of code and collaboration
diff --git a/pkg/workflow/expression_extraction.go b/pkg/workflow/expression_extraction.go
index ce3156c3a9d..00b92265f1c 100644
--- a/pkg/workflow/expression_extraction.go
+++ b/pkg/workflow/expression_extraction.go
@@ -199,6 +199,14 @@ func transformActivationOutputs(expr string) string {
 // experimentNameRegex matches experiments.<name> expressions where name is a simple identifier.
 var experimentNameRegex = regexp.MustCompile(`^experiments\.([a-zA-Z_][a-zA-Z0-9_]*)$`)
 
+// experimentComparisonRegex matches experiments.<name> followed by a comparison operator and
+// a quoted string value, e.g. `experiments.prompt_style == "concise"` or
+// `experiments.prompt_style !== "detailed"`. The value must be enclosed in double quotes
+// with no embedded quotes. It captures:
+//   - group 1: the experiment name
+//   - group 2: the remainder of the expression (operator + quoted value), verbatim
+var experimentComparisonRegex = regexp.MustCompile(`^experiments\.([a-zA-Z_][a-zA-Z0-9_]*)([ \t]*(?:!==?|===?)[ \t]*"[^"]*"[ \t]*)$`)
+
 // ExperimentEnvVarName returns the env-var name used for the given experiment.
 // The name is uppercased; hyphens are converted to underscores; all other characters
 // that are not A-Z, 0-9, or underscore are dropped (not replaced).
@@ -209,17 +217,28 @@ func ExperimentEnvVarName(experimentName string) string {
 }
 
 // transformExperimentsExpression detects expressions of the form "experiments.<name>"
-// and rewrites them to "steps.pick-experiment.outputs.<name>" so that the placeholder
-// substitution step reads the value from the pick_experiment step output.
-// This is used for ${{ experiments.name }} expressions that appear directly in the prompt body
-// (mostly relevant in inline mode; in runtime-import mode the template conditional
-// {{#if experiments.name}} path is handled separately via ExperimentExpressionMappings).
+// (and the comparison form `experiments.<name> == "value"`) and rewrites them so that the
+// placeholder substitution step reads the value from the pick_experiment step output.
+//
+// Simple form:     experiments.name          → steps.pick-experiment.outputs.name
+// Comparison form: experiments.name == "v"  → steps.pick-experiment.outputs.name == "v"
+//
+// This is used for ${{ experiments.name }} and ${{ experiments.name == "value" }} expressions
+// that appear directly in the prompt body (mostly relevant in inline mode; in runtime-import
+// mode the {{#if experiments.name == "value"}} conditional is handled by interpolate_prompt.cjs
+// step 2.5 which substitutes the variant value directly inside the condition tag).
+//
+// Without this transformation, the generated env var would contain an invalid GitHub Actions
+// expression like `${{ experiments.name == "value" }}` where `experiments` is not a real
+// context, causing the expression to always evaluate to false.
 func transformExperimentsExpression(expr string) string {
-	m := experimentNameRegex.FindStringSubmatch(expr)
-	if m == nil {
-		return expr
+	if m := experimentNameRegex.FindStringSubmatch(expr); m != nil {
+		return "steps.pick-experiment.outputs." + m[1]
+	}
+	if m := experimentComparisonRegex.FindStringSubmatch(expr); m != nil {
+		return "steps.pick-experiment.outputs." + m[1] + m[2]
 	}
-	return "steps.pick-experiment.outputs." + m[1]
+	return expr
 }
 
 // simpleIdentifierRegex matches simple JavaScript property access chains like
diff --git a/pkg/workflow/expression_extraction_test.go b/pkg/workflow/expression_extraction_test.go
index a7d2f4d0155..aebb2a2ba1f 100644
--- a/pkg/workflow/expression_extraction_test.go
+++ b/pkg/workflow/expression_extraction_test.go
@@ -61,6 +61,30 @@ func TestExpressionExtractor_ExtractExpressions(t *testing.T) {
 			wantCount:       1,
 			wantExpressions: []string{"steps.pick-experiment.outputs.caveman"},
 		},
+		{
+			name:            "experiments.name == value comparison form gets transformed to step output",
+			markdown:        `{{#if ${{ experiments.prompt_style == "concise" }} }}foo{{/if}}`,
+			wantCount:       1,
+			wantExpressions: []string{`steps.pick-experiment.outputs.prompt_style == "concise"`},
+		},
+		{
+			name:            "experiments.name === value strict-equality form gets transformed to step output",
+			markdown:        `{{#if ${{ experiments.prompt_style === "concise" }} }}foo{{/if}}`,
+			wantCount:       1,
+			wantExpressions: []string{`steps.pick-experiment.outputs.prompt_style === "concise"`},
+		},
+		{
+			name:            "experiments.name != value inequality form gets transformed to step output",
+			markdown:        `{{#if ${{ experiments.reasoning_depth != "multi_candidate" }} }}foo{{/if}}`,
+			wantCount:       1,
+			wantExpressions: []string{`steps.pick-experiment.outputs.reasoning_depth != "multi_candidate"`},
+		},
+		{
+			name:            "experiments.name !== value strict-inequality form gets transformed to step output",
+			markdown:        `{{#if ${{ experiments.reasoning_depth !== "multi_candidate" }} }}foo{{/if}}`,
+			wantCount:       1,
+			wantExpressions: []string{`steps.pick-experiment.outputs.reasoning_depth !== "multi_candidate"`},
+		},
 		{
 			name:            "expression with whitespace",
 			markdown:        "Value: ${{  github.actor  }}",