{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp12_substrate_denial72_20260805","cell_id":"agentic_control_loop_design__computer_science","arm":"CONSTRAINED_HIGH","candidate_id":"agentic_control_loop_design__computer_science__CONSTRAINED_HIGH","decision":"PROPOSAL","abstention_reason":null,"proposal":{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp12_substrate_denial72_20260805","cell_id":"agentic_control_loop_design__computer_science","arm":"CONSTRAINED_HIGH","candidate_id":"agentic_control_loop_design__computer_science__CONSTRAINED_HIGH","proposal_index":1,"version":0,"title":"Self-Powered Thermomechanical Cold-Air Louver for Local Server-Rack Hotspots","problem":"A server rack can experience shifting inlet-air hotspots when server heat loads, fan draw, and exhaust recirculation change. A fixed opening sized for the worst case wastes conditioned air at low load, while a smaller fixed opening may leave a high-load server drawing warm recirculated air. The affected server is expected to regulate its own components but cannot control the temperature or provenance of the air presented at its inlet.","actors":["Server and its internal cooling fans","Cold-aisle supply air","Warm recirculating rack air","Bimetallic sensing-and-actuation element","Spring-return louver and mechanical travel stops","Data-center facilities engineer"],"observable_state":"The directly observable physical state is local server-inlet air temperature, expressed as bimetal deflection and resulting louver angle. Supporting measurements in a test may include inlet temperature, aperture position, plenum pressure, airflow, and component temperature, but those measurements are not required for operation.","consequence":"During inadequate cold-air delivery, the server inlet remains exposed to warmer air, narrowing thermal margin and potentially provoking internal fan acceleration or thermal throttling. Permanently providing maximum cold-air aperture can instead increase bypass flow and disturb rack airflow distribution.","affected_objective":"Keep the selected server inlet within its validated thermal envelope while avoiding a permanently maximal cold-air bypass aperture.","intervention":"Fit a short sealed duct from the cold-aisle face to the selected server inlet with a spring-return louver driven directly by a bimetallic strip exposed to inlet air. Spring preload physically encodes the opening threshold. As inlet temperature rises, differential thermal expansion supplies mechanical work that opens the louver and admits more cold-aisle air; cooling reverses the deflection and reduces the aperture. Minimum and maximum stops bound airflow, a return spring biases the assembly toward a validated fail-open position upon linkage release, and mechanical damping limits hunting. The device uses no motor, processor, network, inferred state, or operator command during operation.","structural_mapping":[{"archetype_element":"Represented Goal","domain_realization":"The desired inlet-temperature region is embodied by bimetal geometry and adjustable spring preload rather than held in software."},{"archetype_element":"World Model","domain_realization":"The material transfer function and linkage embody the bounded causal proposition that excessive local inlet temperature calls for a larger cold-air aperture and that added cold airflow should reduce that temperature."},{"archetype_element":"Observation Channel","domain_realization":"The bimetal is physically exposed to the inlet stream; its differential expansion directly transduces local temperature without electronic sensing or interpretation."},{"archetype_element":"Action Repertoire","domain_realization":"The available actions are the mechanically bounded continuum of louver apertures between validated minimum and maximum stops."},{"archetype_element":"Selection Policy","domain_realization":"The instantaneous force balance among bimetal deflection, spring preload, linkage geometry, and damper resistance selects the aperture."},{"archetype_element":"Legitimate Action Boundary","domain_realization":"Hard stops prevent full blockage and excessive opening; the mechanism has no physical path to change server workload, electrical power, or other rack controls."},{"archetype_element":"Execution Capability","domain_realization":"Thermal expansion supplies the work needed to move the low-friction louver against the return spring."},{"archetype_element":"Effect Feedback Loop","domain_realization":"Opening changes cold-airflow delivery, which changes inlet temperature, which changes bimetal force and the next aperture."},{"archetype_element":"Model Update Rule","domain_realization":"The bimetal and a small thermal mass continuously update the controller's physical state from new inlet conditions; designed thermal inertia and mechanical damping make recent effects influence the next movement without storing digital data."},{"archetype_element":"Override and Escalation Path","domain_realization":"A technician can mechanically latch the louver in its validated fail-open position and remove the module if temperatures, pressure drop, or oscillation exceed test limits."},{"archetype_element":"Agency Health Signal","domain_realization":"Visible alignment marks show whether inlet temperature, bimetal deflection, and louver travel remain physically coupled; they are inspection aids, not the operative feedback channel."}],"mechanism_mapping":[{"mechanism_slug":"agency_loop_map","role":"The embodied loop is goal preload to temperature transduction to bounded aperture selection to airflow change to renewed temperature transduction.","counterfactual_removal":"Breaking the physical return path from louver-induced airflow to the bimetal's inlet temperature converts the device to open-loop motion and removes the adaptive effect."},{"mechanism_slug":"safe_action_menu","role":"Mechanical stops restrict the action repertoire to apertures shown in bench testing not to starve the server or create excessive bypass flow.","counterfactual_removal":"Without the stops, thermomechanical feedback may still move the louver, but unsafe closure or over-opening becomes possible; bounded agency is lost."},{"mechanism_slug":"action_effect_feedback_review","role":"The same material element that initiates an action is re-exposed to the resulting inlet temperature, so the physical consequence directly revises subsequent action.","counterfactual_removal":"Thermally isolating the element from post-action inlet air prevents action effects from correcting later louver position and removes the essential closed-loop regulation."},{"mechanism_slug":"graduated_autonomy_ramp","role":"Linkage geometry makes small temperature excursions produce limited travel and progressively larger excursions produce greater opening, up to a hard limit.","counterfactual_removal":"Replacing graduated travel with an unconstrained snap to either extreme preserves some thermostat behavior but removes proportional local modulation and increases hunting and airflow-distribution risk."}],"causal_chain":["Changing compute load or rack recirculation raises air temperature at the selected server inlet.","The exposed bimetal develops greater differential-expansion curvature.","Bimetal force overcomes more of the calibrated spring force.","The linkage increases louver aperture within mechanical stops.","A larger cold-aisle airflow fraction enters the sealed inlet duct.","The warmer recirculated-air fraction at the inlet decreases.","Inlet cooling reduces bimetal curvature and therefore reduces further opening or begins closure.","Repeated physical observation of the changed temperature settles the louver near a load-dependent force balance."],"baseline":"Use a fixed perforated blanking panel or manually selected aperture sized as a compromise between hotspot protection and cold-air bypass. This baseline cannot change its aperture when local thermal conditions change.","nearest_rivals":["A larger fixed vent or permanently open duct, which is simple but cannot reduce bypass flow during cooler conditions","Rack blanking panels, brushes, and passive containment baffles that reduce recirculation without locally modulating cold-air admission","Increasing server or room fan speed, which uses powered actuation and may alter airflow beyond the affected inlet","An electronically sensed motorized damper or building-control loop, which offers configurable control but depends on sensors, power, computation, and control software","A replaceable phase-change or wax-actuated thermal vent, which is also non-electronic but may have different response time, hysteresis, and service behavior"],"remaining_contrastive_claim":"Subject to bench validation, the proposed device differs from a fixed aperture by physically coupling local inlet temperature to bounded cold-air admission, and differs from electronic dampers by retaining that coupling through material deformation and mechanical feedback alone. No claim of superiority, novelty, prevalence, demand, or effect magnitude is made.","authority_safety":{"decision_authority":"A data-center facilities or thermal engineer retains authority to select the target inlet, validate the airflow envelope, set mechanical stops and preload, authorize installation, and remove the device. The mechanism is delegated authority only to vary its own aperture inside those stops.","authorized_first_step":"Construct one instrumented bench module and test it on a non-production rack airflow mock-up under guarded, reversible heat-load and recirculation steps.","excluded_actions":["Installation on production servers before thermal and pressure-drop validation","Permitting the louver to reach an unvalidated fully closed position","Coupling the mechanism to workload control, server power, network access, or other rack actuators","Treating visible marks, logging, alerts, or operator response as necessary to produce the cooling effect","Using the prototype in a fire-rated or smoke-control boundary without separate compliance review"],"halt_rollback":"Halt a trial if inlet or component temperature crosses the mock server's predetermined test limit, airflow falls below its minimum, the louver oscillates persistently, the linkage binds, or the fail-open return fails. Remove heat, latch the damper open, restore the fixed-duct baseline, and remove the module before diagnosis."},"negative_tests":{"strongest_counterevidence":"The strongest counterevidence would be a repeatable finding that greater louver opening does not reduce the target inlet's warm-air fraction or temperature because rack pressure fields redirect flow, while the device instead deprives neighboring inlets or increases recirculation.","problem_falsifier":"Across representative load and recirculation steps, the target inlet does not experience meaningful state variation addressable by local aperture changes, or a fixed containment correction removes the variation without adaptive modulation.","intervention_falsifier":"In a rack mock-up, the thermomechanical module fails to move within the required thermal-response window, fails to make aperture track inlet temperature in the intended direction, produces hunting or binding, or does not improve the target inlet's temperature trajectory relative to matched fixed apertures without causing unacceptable neighboring airflow changes.","risks":["Thermal lag may cause delayed response or oscillation.","A broken or contaminated linkage may stick at an unsafe aperture.","The bimetal calibration may drift with fatigue, corrosion, or mounting stress.","Additional pressure drop may reduce server-fan airflow.","Redirected cold air may worsen temperatures at neighboring servers.","The local temperature-to-airflow causal assumption may fail under some rack pressure regimes.","A fail-open configuration may consume excess conditioned air after a fault.","Combustibility, smoke movement, vibration, acoustic, and maintenance constraints may make a candidate material or placement unsuitable."]},"next_evidence_step":"On a non-production airflow mock-up, compare the prototype with minimum, midpoint, and maximum fixed-aperture controls across a bounded matrix of heat-load and recirculation steps. Record inlet and neighboring-inlet temperatures, airflow, pressure drop, louver travel, settling time, oscillation, and fail-open behavior. Repeat the direction-changing steps to expose hysteresis and binding. Proceed only if the physical action-effect direction is repeatable, stops and fail-open behavior work, and no predefined thermal or airflow limit is crossed.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Not assessed against other proposals because runtime isolation forbids their inspection; this candidate is defined solely by its passive thermomechanical airflow loop.","revision_record":{"parent_version":null,"progress_targets_addressed":["Construct one causally complete candidate under the binding physical-substrate constraint.","Preserve goal, observation, bounded action, execution, feedback, and state-update structure.","Specify serious physical and electronic rivals, safeguards, falsifiers, and a bounded evidence step."],"conceptual_changes":["Initial version; no parent proposal.","Realized agentic control as an embodied thermomechanical loop rather than a software agent, dashboard, or governance process."],"operational_changes":["Initial version; specified mechanical stops, damping, sealed ducting, and fail-open rollback."],"evidence_changes":["Initial version; prior art remains unsearched and the first evidence step is restricted to a rack mock-up."],"claim_changes":["Limited the contrastive claim to causal architecture and explicitly withheld novelty, superiority, prevalence, demand, and effect-size claims."]},"substrate_contract":{"primary_allowed_process":"PHYSICAL_MATERIAL","counterfactual_independence":"If all software, algorithms, databases, dashboards, reporting, incentives, permissions, reviews, workflows, training, and procedural enforcement are removed, inlet heat still bends the bimetal, the linkage still changes the louver aperture, airflow still changes inlet temperature, and that temperature still feeds back through the same material element. The essential adaptive cooling effect therefore survives removal of every forbidden wrapper.","forbidden_channel_audit":"No processor, inferred state, electronic sensor, motor, network, report, human response, incentive, or policy closes the operative loop. Optional test instrumentation only measures performance during evaluation. Engineering authorization, calibration, maintenance, and visual alignment marks support safe deployment but do not cause aperture modulation or cooling. The causal path is exclusively temperature to differential material expansion to mechanical displacement to spatial airflow change to temperature feedback."}}}