Archive of Christian Ullrich

AI Reflex OS

GenAI and Military Power Reflex Area

2026-09-25

Table of Contents

• Assessing GenAI-Enabled Military Advantage
  • Assessing Military Capabilities GenAI Does Not Materially Change
  • Connecting Administrative Productivity to Readiness
  • Connecting Staff Work Improvements to Operational Effect
  • Distinguishing Local Efficiency From Military Advantage
  • Assessing High-Frequency GenAI Uses With Cumulative Effects
  • Evaluating GenAI as a Substitute for Scarce Capacity
  • Evaluating GenAI as a Multiplier of Existing Capability
  • Assessing Whether GenAI Changes the Real Constraint
  • Assessing Military Effect Beyond GenAI Activity Measures
  • Assessing GenAI Use With Unclear Strategic Value
• Assessing Relative Capability and Adaptation
  • Comparing Forces With Similar GenAI Access
  • Assessing an Adversary's GenAI Exploitation Capability
  • Comparing Adaptation Speed Across Competitors
  • Distinguishing Model Superiority From Institutional Superiority
  • Assessing Whether a GenAI Lead Is Temporary or Durable
  • Comparing Absolute Improvement With Relative Position
  • Assessing Asymmetric Gains for Smaller or Less-Resourced Actors
  • Identifying Traditional Advantages Weakened by GenAI Diffusion
  • Assessing Whether GenAI Narrows or Widens Existing Capability Gaps
  • Assessing Relative Military Disadvantage From Slower GenAI Adaptation
• Managing Organizational Tempo and Shifting Bottlenecks
  • Accelerating One Stage Without Accelerating the Whole Process
  • Identifying the New Bottleneck After GenAI Speeds Up Work
  • Distinguishing Cognitive Delay From Institutional Delay
  • Assessing Approval Structures After Preparation Time Falls
  • Managing Handoffs After Upstream Work Accelerates
  • Evaluating Whether Faster Staff Work Improves Decision Tempo
  • Maintaining Decision Quality as Decision Cycles Compress
  • Handling More Options Than Leaders Can Evaluate
  • Identifying Physical Constraints After Cognitive Work Accelerates
  • Reassessing Tempo When Functions Accelerate at Different Rates
• Assessing Compounding Learning Advantage and Path Dependence
  • Comparing Accumulated GenAI Experience Across Forces
  • Assessing Whether Early Experience Is Creating Durable Advantage
  • Identifying Tacit GenAI Competence That Cannot Be Acquired Quickly
  • Distinguishing Transferable Lessons From Local Experience
  • Assessing Whether a Latecomer Can Leapfrog Earlier Users
  • Assessing Lock-In From Early GenAI Choices
  • Assessing Whether Local GenAI Experience Produces Force-Level Advantage
  • Estimating the Cost of Delayed Learning
  • Identifying Path-Dependent Capability Gaps
  • Reassessing Early Advantages as GenAI Capabilities Change
• Managing Cognitive Abundance and Scarce Attention
  • Handling More Analysis Than Leaders Can Consume
  • Prioritizing Attention When Cognitive Production Becomes Cheap
  • Assessing Military Consequences of Freed Staff Capacity
  • Identifying New Scarce Resources After GenAI Expansion
  • Review Capacity Becoming Scarce as GenAI Output Expands
  • Allocating Scarce Command and Expert Judgment Across GenAI-Expanded Work
  • Reassessing Reporting When Summarization Becomes Cheap
  • Staff Work Expanding Because GenAI Makes Cognitive Production Cheap
  • Managing Convergence in GenAI-Assisted Staff Analysis
  • Recognizing Organizational Structures Built Around Scarce Cognitive Capacity
• Redefining Expertise, Staffs, and Command
  • Assessing the Shift From Production Expertise Toward Evaluation Expertise
  • Preserving Expertise Needed to Challenge GenAI
  • Assessing How GenAI Changes the Functions of Military Staffs
  • Reconsidering Headquarters Size and Composition
  • Evaluating Whether GenAI Should Expand Span of Control
  • Managing GenAI-Enabled Micromanagement
  • Reassessing How Authority Is Distributed Across Command Levels
  • Maintaining Command Responsibility in GenAI-Supported Decisions
  • Calibrating Command Reliance on GenAI Decision Support
  • Preserving Command Judgment Through Routine GenAI Use
• Managing GenAI-Driven Capability Evolution
  • Reassessing a Fielded Capability After a Major GenAI Capability Shift
  • Deciding Whether to Retain, Integrate, Adapt, or Redesign an Existing System
  • Managing Requirements That Change Faster Than Acquisition Cycles
  • Reassessing Doctrine After GenAI Changes Available Options
  • Assessing When Training Baselines Become Obsolete Faster Than Training Cycles
  • Managing Continuous Model and Software Change in Fielded Capability
  • Retesting Capabilities After Model or Data Changes
  • Coordinating Doctrine, Training, Systems, and Organization as One Changes
  • Managing Diverging GenAI Capability Baselines Across the Force
  • Assessing When Existing Systems or Processes Become Capability Constraints
• Managing GenAI Ecosystem Dependence and Military Autonomy
  • Relying on Commercial GenAI Providers for Military Work
  • Assessing Dependence on Foreign Models or Infrastructure
  • Evaluating Cloud and Compute Dependence
  • Managing Semiconductor and Hardware Supply Dependence
  • Responding When Commercial GenAI Evolves Faster Than Military Absorption
  • Preparing for Provider Policy, Pricing, or Access Changes
  • Deciding Which GenAI Capabilities Require Assured Access
  • Balancing Commercial Innovation With Military Autonomy
  • Assessing Switching Costs Between GenAI Providers
  • Identifying Hidden Dependencies Across the GenAI Ecosystem
• Balancing Collective GenAI Capability and Common-Mode Risk
  • Operating With Uneven GenAI Capability Across Partners
  • Sharing GenAI Services Across Allied Organizations
  • Assessing Whether National GenAI Advantage Transfers to Collective Operations
  • Maintaining Collective Tempo Across Different GenAI Maturity Levels
  • Balancing Common GenAI Platforms With National Flexibility
  • Assessing Whether Data Asymmetries Limit Collective GenAI Capability
  • Assessing Common-Model Dependence Across Partners
  • Assessing Cognitive Diversity With Shared GenAI Systems
  • Evaluating Correlated Failure Across Shared GenAI Infrastructure
  • Responding When One Partner's GenAI Dependency Constrains the Group
• Maintaining Military Capability Under GenAI Degradation
  • Operating After Loss of a Mission-Critical GenAI Service
  • Working Through Degraded Connectivity to GenAI
  • Responding to Compromised or Unreliable GenAI Outputs
  • Switching to Alternative GenAI Support Under Constraint
  • Falling Back to Human-Only Workflows
  • Maintaining Critical Functions With Reduced GenAI Support
  • Preserving Human Competence for Degraded Operations
  • Testing Mission Capability Without Preferred GenAI Support
  • Assessing GenAI Dependence Before a Failure
  • Recovering From a Shared GenAI Ecosystem Outage
• Anticipating Strategic Surprise From GenAI-Enabled Adaptation
  • Detecting Adversary Capability Gains Without New Platforms
  • Reassessing Intelligence Indicators for GenAI-Enabled Change
  • Recognizing Military Effects From Commercial GenAI Diffusion
  • Assessing Sudden GenAI Uplift in Smaller Actors or Proxies
  • Responding to Unexpected Changes in Adversary Decision Tempo
  • Recognizing When Institutional Adaptation Outpaces Existing Assumptions
  • Distinguishing Genuine Competitive Urgency From GenAI Arms-Race Rhetoric
  • Identifying When a GenAI Adaptation Gap Becomes a Readiness Gap
  • Assessing the Significance of an Unexpected GenAI Capability Demonstration
  • Reassessing the Time Available to Adapt After GenAI Capability Shifts

Assessing GenAI-Enabled Military Advantage

Assessing Military Capabilities GenAI Does Not Materially Change

• Which military outcomes would remain constrained even if GenAI made the relevant cognitive work dramatically faster?
• Is the binding constraint physical capacity, logistics, geography, infrastructure, authorities, trained personnel, industrial capacity, or something else?
• Where could GenAI improve understanding of a constraint without materially changing the constraint itself?
• Which parts of the capability depend on embodied skill, materiel, physical presence, or infrastructure that GenAI cannot supply?
• Would better analysis change what the force can actually generate, move, maintain, protect, or sustain?
• Could attention to GenAI divert resources from a more consequential noncognitive capability need?
• What evidence would show that further GenAI improvement is unlikely to change this military capability materially?

Connecting Administrative Productivity to Readiness

• Which administrative activities currently delay training, maintenance, personnel availability, deployment, recovery, or force generation?
• Is the GenAI-supported administrative step actually on the critical path to a readiness outcome?
• What happens to the personnel time or organizational capacity released when the administrative work becomes faster?
• Can the downstream readiness activity absorb and use the capacity that GenAI releases?
• Which specific readiness delay could disappear if this administrative work were completed earlier?
• What readiness measure should improve if the productivity gain is genuinely propagating into military capability?
• Under what conditions could substantial administrative efficiency leave readiness almost unchanged?

Connecting Staff Work Improvements to Operational Effect

• Which staff output materially affects a command decision, operational preparation, coordination, or execution?
• What exactly becomes faster, better, broader, or more reliable when GenAI supports that staff work?
• Does the improved output reach the decision-maker or operator early enough to change what they can do?
• Which approval, coordination, authority, resource, or execution constraint could prevent the staff gain from propagating further?
• Is GenAI reducing the commander's workload or simply allowing the staff to send more material upward?
• What end-to-end measure would show that the operational process improved rather than just the staff activity?
• What evidence would connect the staff-level improvement to a consequential change in military performance?

Distinguishing Local Efficiency From Military Advantage

• What specific task or process has become faster, cheaper, easier, or less labor-intensive?
• Which larger military process or capability is supposed to benefit from that local improvement?
• Does the gain relax a binding constraint or merely reduce effort in a part of the system that was not limiting performance?
• Can downstream functions absorb the additional speed, volume, or capacity without creating a new bottleneck?
• Would the improvement still create relative advantage if relevant competitors received a similar productivity gain?
• What new verification, coordination, integration, or support costs accompany the apparent efficiency?
• What evidence would justify calling the result military advantage rather than useful local productivity?

Assessing High-Frequency GenAI Uses With Cumulative Effects

• Which GenAI-supported activities occur often enough for modest improvements to accumulate at meaningful scale?
• How much time, delay, specialist effort, or cognitive capacity does the activity consume across its full annual or operational volume?
• Where does the capacity released across thousands of repetitions actually go?
• Do verification effort, errors, rework, or additional output accumulate alongside the productivity gain?
• Can downstream functions use the increased throughput without becoming the new constraint?
• Does repeated use create additional learning and capability, or simply expand the amount of work the organization expects?
• At what scale would the cumulative effect become visible in readiness, capacity, tempo, resilience, or another military outcome?

Evaluating GenAI as a Substitute for Scarce Capacity

• What scarce human or organizational capacity is GenAI expected to replace or extend?
• Which parts of that capacity are routine enough for GenAI to perform adequately, and which still depend on human expertise or judgment?
• What level of output quality is sufficient for GenAI to count as a practical substitute in this military function?
• Does substitution remove the scarcity, or does it create a new bottleneck in review, supervision, approval, or integration?
• What valuable qualities would disappear if the scarce human capacity also provides trust, contextual knowledge, accountability, or relationships?
• How resilient would the function remain if the GenAI substitute became unavailable or unreliable?
• What evidence would show that effective military capacity increased after accounting for supervision, verification, and fallback requirements?

Evaluating GenAI as a Multiplier of Existing Capability

• Which existing military capability becomes more effective when GenAI is added to it?
• Is GenAI multiplying speed, scale, precision, accessibility, learning, adaptability, or some other useful characteristic?
• Which complementary assets must already exist for that multiplier effect to occur?
• Would a weak underlying capability remain weak even with substantially better GenAI support?
• What new constraint becomes important once the existing capability is amplified?
• Does the multiplier introduce dependencies that could reduce resilience or freedom of action?
• What observable military effect would show that GenAI is amplifying capability rather than merely increasing cognitive output?

Assessing Whether GenAI Changes the Real Constraint

• What currently limits the military outcome we are trying to improve?
• What evidence shows that this is the binding constraint rather than simply a visible or inconvenient part of the process?
• If GenAI made the targeted cognitive work nearly instantaneous, what would become the next source of delay or scarcity?
• Is the real constraint cognitive, institutional, physical, logistical, legal, organizational, or something else?
• Does GenAI actually relax the constraint, or does it mainly provide better information about it?
• What complementary change would be necessary for the GenAI improvement to affect the final military outcome?
• How should priorities change if the GenAI intervention is acting on a nonbinding part of the system?

Assessing Military Effect Beyond GenAI Activity Measures

• What military outcome is the GenAI activity ultimately supposed to influence?
• Which current measures describe usage or output rather than changes in military capability?
• What process-level measure would show whether the GenAI gain propagated beyond the immediate task?
• What readiness, capacity, tempo, resilience, quality, or mission measure should eventually change?
• How could we distinguish the effect of GenAI from other changes occurring in the same period?
• What negative effects, verification costs, rework, dependencies, or additional burdens should be measured alongside the benefits?
• What evidence threshold would justify continuing, scaling, changing, or stopping the GenAI-supported activity?

Assessing GenAI Use With Unclear Strategic Value

• What underlying problem is this GenAI use actually solving?
• Which military outcome would be materially different if the use disappeared?
• Is the main benefit convenience, productivity, learning, experimentation, or genuine military capability?
• What opportunity cost comes from funding, integrating, supporting, and governing the use?
• Do verification, security, integration, coordination, or dependency costs materially reduce the apparent benefit?
• Could the same problem be addressed more effectively through a non-GenAI change?
• What evidence would be needed before treating this use as strategically important?

Assessing Relative Capability and Adaptation

Comparing Forces With Similar GenAI Access

• If two forces can use similarly capable models, what explains any difference in military effect?
• How do their data, infrastructure, integration, expertise, workflows, evaluation, doctrine, and authorities differ?
• Which force is better at turning model capability into recurring military performance rather than isolated experimentation?
• Does one force learn faster from operational use and feed those lessons back into subsequent adaptation?
• Which complementary military capabilities make the same GenAI more valuable to one force than the other?
• What advantages would remain if both forces received exactly the same model tomorrow?
• What evidence would demonstrate a meaningful exploitation gap rather than a difference in technology access?

Assessing an Adversary's GenAI Exploitation Capability

• What evidence shows the adversary moving beyond access, demonstrations, or isolated pilots into sustained military use?
• Is GenAI embedded in recurring workflows, exercises, training, software, decision support, or capability development?
• What data, infrastructure, expertise, evaluation, governance, and support structures sustain the apparent capability?
• Are there observable changes in military behavior or performance that are consistent with successful exploitation?
• How well could the capability continue under disruption, degraded connectivity, personnel loss, or provider change?
• Which important parts of the exploitation capability could remain hidden from external observation?
• What alternative explanations should be tested before attributing an observed military improvement to GenAI?

Comparing Adaptation Speed Across Competitors

• How quickly does each force recognize a GenAI development that is relevant to military capability?
• How long does each force take to move from recognition to experimentation, validation, integration, and recurring use?
• Where do procurement, security, authority, data, training, or organizational processes accelerate or slow adaptation?
• How quickly do successful local practices spread beyond the original unit or technical team?
• Does operational experience feed back into software, doctrine, training, and subsequent experimentation?
• Is each organization learning quickly enough to keep pace with changes in the underlying technology?
• What evidence would show that differences in adaptation speed are becoming militarily consequential?

Distinguishing Model Superiority From Institutional Superiority

• How much of the observed advantage comes from access to a better model rather than better exploitation of available models?
• Would the advantage survive if the competitor gained access to the same model?
• Which data, workflows, integration, evaluation, expertise, and organizational routines are producing superior results?
• Is the force able to replace models without losing the surrounding military capability?
• Which elements of the advantage can competitors acquire commercially and which require accumulated institutional learning?
• Could reliance on a superior current model conceal weaknesses in the organization's underlying exploitation capability?
• What evidence would justify describing the advantage as institutional rather than primarily technological?

Assessing Whether a GenAI Lead Is Temporary or Durable

• What is the actual source of the current GenAI-related lead?
• How easily could competitors acquire, imitate, or substitute the technology creating that lead?
• Does the advantage depend on accumulated data, integration, evaluation, expertise, trusted workflows, or organizational experience?
• Is the leading force using its current position to learn and adapt faster than competitors?
• Are early choices creating lock-in that could turn today's advantage into tomorrow's constraint?
• What technological, commercial, organizational, or geopolitical change could quickly erase the lead?
• Which parts of the advantage would remain valuable after the current models and providers were replaced?

Comparing Absolute Improvement With Relative Position

• How much has our military capability improved over the relevant period?
• How much have the most relevant competitors improved over the same period?
• Are we using comparable military outcomes and timeframes when assessing both sides?
• Could our absolute performance be improving while our relative position is deteriorating?
• Are differences driven by model access, exploitation quality, complementary assets, or adaptation speed?
• Which military capability gap matters operationally rather than simply appearing large in technical metrics?
• What evidence would show that our improvement has strengthened our competitive position rather than only our internal baseline?

Assessing Asymmetric Gains for Smaller or Less-Resourced Actors

• Which scarcity facing the smaller actor can GenAI reduce most sharply?
• Does GenAI substitute for staff capacity, language capability, software skill, specialist knowledge, analysis, or another scarce cognitive resource?
• What military or security activity becomes newly feasible or substantially easier as a result?
• Which disadvantages remain because they depend on physical scale, logistics, industry, sensors, trained forces, or secure infrastructure?
• Does the actor receive a large percentage gain without gaining a correspondingly large absolute military capability?
• What complementary assets from a sponsor, partner, or commercial provider would amplify the GenAI effect?
• Under what conditions would the uplift become important enough to alter the actor's operational or strategic significance?

Identifying Traditional Advantages Weakened by GenAI Diffusion

• Which existing military advantages depend partly on expensive or scarce cognitive production?
• What happens to those advantages if analysis, translation, coding, synthesis, or planning support becomes widely accessible?
• Which specialist capabilities could competitors now approximate without building the same large human organization?
• Which scale advantages remain robust because they depend on physical, institutional, industrial, or operational resources?
• What new scarce resource becomes more important as cognitive production becomes cheaper?
• How quickly could commercial diffusion erode the advantage in practice rather than in theory?
• Which traditional strengths should be reinforced, complemented, or redefined as their underlying scarcity changes?

Assessing Whether GenAI Narrows or Widens Existing Capability Gaps

• Which specific capability gap is GenAI affecting?
• Which parts of that gap arise from cognitive scarcity and which arise from physical, institutional, industrial, or human differences?
• Does GenAI allow the weaker actor to compensate for an existing disadvantage?
• Does the stronger actor possess complementary assets that let it exploit the same technology even more effectively?
• Could GenAI narrow one dimension of the gap while widening another?
• How is the gap changing over time as models become cheaper, more capable, and more widely available?
• What evidence would show whether the overall military difference is converging, diverging, or changing form?

Assessing Relative Military Disadvantage From Slower GenAI Adaptation

• At which stage of adaptation are we moving more slowly than the relevant competitor?
• Is the lag caused by technology access, experimentation, evaluation, integration, authorities, procurement, training, or institutional learning?
• Which military capability is beginning to reflect the difference in adaptation speed?
• What experience and organizational knowledge is the competitor accumulating while the gap persists?
• Could the difference still be closed mainly by acquiring technology, or has it become an institutional capability gap?
• How quickly could we close the gap if operational circumstances suddenly made it urgent?
• What evidence would justify treating slower adaptation as a military disadvantage rather than a general modernization concern?

Managing Organizational Tempo and Shifting Bottlenecks

Accelerating One Stage Without Accelerating the Whole Process

• Which exact stage of the military process has become faster because of GenAI?
• Was that stage on the critical path before it accelerated?
• What waiting time remains before and after the faster stage?
• Is work now accumulating in a downstream queue that cannot absorb the increased arrival rate?
• Has total elapsed time to the military decision, output, or effect actually decreased?
• What new review, coordination, approval, or execution burden has appeared because upstream production is faster?
• What else would have to change before the local acceleration becomes genuine end-to-end tempo improvement?

Identifying the New Bottleneck After GenAI Speeds Up Work

• What constraint limited the process before GenAI changed the workflow?
• What resource or stage now determines throughput after the original constraint has been reduced?
• Is the new bottleneck human attention, expert review, authority, coordination, implementation, logistics, or physical capacity?
• What queue, backlog, delay, or utilization pattern shows that the bottleneck has moved?
• Are existing performance measures still focused on a constraint that GenAI has already removed?
• Is the new bottleneck an unnecessary friction or a deliberate control that still serves an important function?
• What organizational response would address the new constraint without simply moving the problem again?

Distinguishing Cognitive Delay From Institutional Delay

• How much of the total elapsed time is active cognitive work and how much is waiting?
• Which delays come from analysis, drafting, synthesis, retrieval, coding, or other information processing?
• Which delays come from authorities, approvals, security processes, policy, meetings, coordination, procurement, or organizational routines?
• Does GenAI act on the source of delay that actually governs the end-to-end process?
• Which institutional delays become visible only after cognitive work gets faster?
• Which remaining delays protect accountability, safety, coordination, or judgment and should not simply be removed?
• What non-GenAI change would be necessary if institutional delay now dominates the process?

Assessing Approval Structures After Preparation Time Falls

• Why does each major approval exist in the current process?
• Does faster preparation increase the number or frequency of matters reaching the same approver?
• Which approvals still require scarce judgment, authority, risk acceptance, or cross-functional coordination?
• Which decisions could be delegated fully or within predefined thresholds?
• Could routine matters move to notification, sampling, or exception-based review instead of prior approval?
• How would accountability remain clear if decision authority moved closer to the point of action?
• What evidence would show that an approval change improved tempo without producing unacceptable decision or control failures?

Managing Handoffs After Upstream Work Accelerates

• Which receiving function now gets work sooner or in greater volume because GenAI accelerated the upstream stage?
• Can the receiving function process the new arrival rate without creating delay, rework, or quality loss?
• What context, evidence, provenance, assumptions, or uncertainty need to accompany the handoff?
• Is faster upstream production transferring hidden verification or interpretation work to the downstream function?
• How do different operating tempos on each side of the handoff affect coordination and synchronization?
• Has responsibility become less clear as work moves more frequently between people or functions?
• What redesign of the interface would allow the upstream gain to propagate without overwhelming the receiver?

Evaluating Whether Faster Staff Work Improves Decision Tempo

• Which part of decision preparation has actually become faster?
• Does useful information reach the commander earlier, or does the staff mainly produce more material within the same decision cycle?
• Has the time from relevant change in the situation to a sufficiently good decision decreased?
• Which authority, coordination, or judgment delays remain unchanged after staff preparation accelerates?
• Is the commander gaining more time for judgment or losing attention to increased staff output?
• Can downstream units act soon enough for the earlier decision to create operational value?
• What evidence would show improved command decision tempo rather than increased staff throughput?

Maintaining Decision Quality as Decision Cycles Compress

• Which elements of decision quality are most vulnerable when the cycle becomes shorter?
• Are evidence checks, dissent, alternative hypotheses, contextual understanding, or expert challenge being compressed disproportionately?
• Which decisions are reversible enough to tolerate greater speed and which still require deliberate judgment?
• Is fluent GenAI-supported preparation creating confidence that exceeds the strength of the underlying evidence?
• What warning signs would show that faster cycles are increasing error, rework, surprise, or poor decisions?
• Which safeguards can be targeted to high-consequence or uncertain decisions rather than applied uniformly?
• How should decision quality and decision speed be measured together so one is not improved by quietly sacrificing the other?

Handling More Options Than Leaders Can Evaluate

• Is GenAI producing genuinely different alternatives or mostly superficial variations of the same basic course?
• At what point does generating another option add little additional decision value?
• Which criteria should determine which alternatives deserve command attention?
• How much expert evaluation is required before an option is mature enough to reach the decision-maker?
• Could abundant options obscure the decisive tradeoffs rather than clarify them?
• Is GenAI being used to filter options in a way that could suppress unconventional but important alternatives?
• What practical limit on option generation would protect leadership attention without prematurely narrowing the choice set?

Identifying Physical Constraints After Cognitive Work Accelerates

• Which physical activity now determines the pace after planning, analysis, or coordination becomes faster?
• Are movement, maintenance, logistics, production, communications, infrastructure, personnel, or training capacity setting the new limit?
• Is faster cognitive work creating expectations that physical execution cannot satisfy?
• Which plans or decisions can now be produced much faster than they can be resourced or implemented?
• Can GenAI materially reduce the physical constraint, or can it only optimize and describe it more accurately?
• What buffers or synchronization mechanisms are needed when cognitive and physical processes operate at different speeds?
• What end-to-end measure would show whether cognitive acceleration changed actual military execution?

Reassessing Tempo When Functions Accelerate at Different Rates

• Which connected military functions are accelerating most because of GenAI?
• Which functions remain constrained by slower human, institutional, coalition, or physical processes?
• At which interfaces do those differing rates create queues, rework, overload, or coordination failure?
• Which activities require a common synchronization point even if their internal speeds differ?
• Could a faster function overwhelm a slower supporting function or partner?
• Where can processes be decoupled or buffered so every function does not have to move at the same speed?
• What measure best reflects collective tempo when different parts of the system are accelerating unevenly?

Assessing Compounding Learning Advantage and Path Dependence

Comparing Accumulated GenAI Experience Across Forces

• How much practical GenAI experience has each force accumulated in recurring military work rather than demonstrations?
• What failures, edge cases, verification problems, and successful operating patterns has each organization actually encountered?
• Is that experience concentrated in individuals or retained through training, infrastructure, evaluation, procedures, and institutional memory?
• How effectively does operational feedback change subsequent GenAI use?
• Which force can evaluate and integrate a new model faster because of what it has already learned?
• What important experiential differences could remain difficult to observe from outside the organization?
• What evidence would show that accumulated experience is becoming a meaningful military capability gap?

Assessing Whether Early Experience Is Creating Durable Advantage

• What has the early adopter learned that later access to a better model would not automatically provide?
• Has early use produced durable assets such as data, evaluation methods, infrastructure, workflows, training, or organizational routines?
• Does accumulated experience make each subsequent experiment cheaper, faster, or more informative?
• Would the advantage survive replacement of the models and interfaces that originally produced it?
• Could a later entrant copy the explicit lessons while avoiding the early adopter's mistakes?
• Are early commitments creating enough lock-in to offset some of the learning advantage?
• What evidence would show that the early lead is compounding rather than merely reflecting an earlier start date?

Identifying Tacit GenAI Competence That Cannot Be Acquired Quickly

• What do experienced GenAI practitioners recognize that written guidance alone does not teach?
• Which failure patterns, contextual cues, or task boundaries become visible only through repeated real-world use?
• How much does effective GenAI judgment depend on combining domain expertise with practical model experience?
• Which aspects of the competence can be taught explicitly and which require exposure to successes, failures, and edge cases?
• Which roles would create the largest capability loss if experienced practitioners suddenly left?
• How quickly could a replacement develop equivalent judgment under realistic working conditions?
• What institutional arrangements help preserve the value of tacit competence without making the organization dependent on a few individuals?

Distinguishing Transferable Lessons From Local Experience

• Which part of a successful GenAI practice reflects a general mechanism and which part depends on the local implementation?
• How dependent is the lesson on a specific mission, model, dataset, classification environment, workforce, or workflow?
• Should the lesson be treated as portable, adaptable with local redesign, or genuinely local?
• What evidence exists that the practice works outside the unit or setting where it originated?
• Could copying the visible method reproduce the form of the practice without reproducing the reason it worked?
• What should remain experimental rather than becoming force-wide guidance or doctrine?
• How can the organization transfer the underlying principle while allowing implementation to vary with local conditions?

Assessing Whether a Latecomer Can Leapfrog Earlier Users

• Which parts of the early adopter's historical learning path can the latecomer now skip?
• Can mature commercial tools or publicly known practices eliminate mistakes that early users had to discover themselves?
• Which capabilities still require direct experience and cannot be compressed simply by acquiring better technology?
• Does the latecomer possess complementary assets that could let it integrate newer GenAI faster than the incumbent?
• Are earlier adopters constrained by legacy providers, architectures, workflows, or assumptions?
• Could a cleaner technical and organizational starting point outweigh the early adopter's accumulated experience?
• What would show that the latecomer has actually leapfrogged rather than merely adopted newer technology?

Assessing Lock-In From Early GenAI Choices

• Which provider, architecture, data, workflow, evaluation, or skill choices are becoming costly to reverse?
• What future options are being narrowed because current systems depend on those early decisions?
• How much operational disruption would switching now require?
• Are the benefits of deep integration large enough to justify the reduced flexibility?
• Could a model or workflow optimized for today's limitations become a constraint as GenAI capabilities change?
• Which interfaces, data structures, and evaluation assets should remain deliberately portable?
• Are we developing conceptual lock-in by treating early assumptions about GenAI as permanent truths?

Assessing Whether Local GenAI Experience Produces Force-Level Advantage

• What useful GenAI experience is emerging in individual units, staffs, or specialist communities?
• Is that experience being captured with enough evidence to understand why the practice succeeded or failed?
• Which lessons can be converted into reusable infrastructure, evaluation methods, training, or operating guidance?
• What prevents strong local practices from reaching other parts of the force that could use them?
• Are force-level mechanisms spreading mature lessons without prematurely standardizing immature practices?
• Does local experimentation continue to generate useful variation after some common practices are adopted?
• What evidence would show that distributed experimentation is compounding into force-wide military capability?

Estimating the Cost of Delayed Learning

• What practical GenAI experience is the force not accumulating while adaptation is delayed?
• Which mistakes, edge cases, integration problems, and operating patterns would still have to be learned later?
• What current military capability is being forgone during the delay?
• Could the delay also reduce future adaptation speed by preventing earlier feedback and organizational learning?
• How much of the missed learning could be imported from partners, vendors, or later technology?
• How long would it take to build the missing competence if the capability suddenly became operationally urgent?
• At what point does the accumulated learning debt outweigh the benefits of waiting for more mature technology?

Identifying Path-Dependent Capability Gaps

• Which current GenAI capability differences reflect accumulated organizational history rather than present-day model access?
• What data, integration, evaluation, expertise, supplier relationships, infrastructure, or trust has developed over time?
• Would equal access to the same models eliminate the gap, or would most of it remain?
• How long would a competitor need to reproduce the path-dependent elements that matter?
• Which historical choices are making current adaptation easier and which are creating legacy constraints?
• Are today's GenAI decisions creating new path dependence that will shape future military options?
• What evidence would show that a capability gap is likely to persist even after technological access equalizes?

Reassessing Early Advantages as GenAI Capabilities Change

• Which early advantages depended on model limitations that no longer exist?
• Have newer GenAI capabilities made previous workarounds, specialist skills, or infrastructure less valuable?
• Which accumulated assets remain useful because they concern data, evaluation, integration, judgment, or organizational learning rather than a particular model?
• Are early users holding onto assumptions that newer technology has already invalidated?
• Can late adopters now achieve comparable results with much less institutional effort?
• What should be retired rather than preserved simply because substantial effort was invested in it?
• Which early investments should continue because they still improve the force's ability to exploit future GenAI generations?

Managing Cognitive Abundance and Scarce Attention

Handling More Analysis Than Leaders Can Consume

• How much more analysis can the staff now produce because of GenAI?
• Is leadership attention increasing at anything close to the same rate?
• Which analyses actually change understanding, decisions, priorities, or action?
• Are important judgments becoming harder to find within the increased volume of plausible material?
• Does additional analysis provide new insight or mainly create greater apparent completeness?
• What filtering should occur before material reaches commanders or senior specialists?
• What measure would show that cognitive abundance is improving decisions rather than merely increasing consumption demands?

Prioritizing Attention When Cognitive Production Becomes Cheap

• Which decisions, risks, uncertainties, and assumptions deserve scarce human attention regardless of how cheaply supporting material can be generated?
• Are people spending time on products mainly because GenAI made them easy to create?
• Which consequences of error justify deeper command or expert examination?
• What should be automated, delegated, sampled, summarized, or ignored so attention remains available for higher-value judgment?
• Can GenAI help prioritize information without becoming the sole authority on what matters?
• Would an explicit information or attention budget prevent low-value cognitive production from expanding indefinitely?
• How should the organization protect attention as a military resource rather than treating it as free?

Assessing Military Consequences of Freed Staff Capacity

• How much staff capacity has GenAI actually released after verification, rework, and coordination costs are included?
• Which important military work is currently constrained by insufficient staff time, attention, or expertise?
• Is released capacity being deliberately redirected or simply absorbed by additional routine workload?
• Has GenAI changed staffing assumptions, or has it mainly increased the amount of work expected from the same staff?
• What previously under-resourced work could now receive sustained attention?
• Which readiness, planning, learning, resilience, coordination, or command outcome should improve if the capacity is being used well?
• What evidence would show that freed staff time has become a military capability gain rather than an internal productivity gain?

Identifying New Scarce Resources After GenAI Expansion

• Which resources are becoming more constrained as cognitive production becomes cheaper?
• Is the emerging scarcity leadership attention, trusted data, expert review, validation, authority, coordination, compute, infrastructure, or execution capacity?
• What evidence shows that the new resource is now limiting military performance?
• Is GenAI increasing demand for the scarce resource faster than the organization can increase supply?
• Are budgets, staffing, and priorities still organized around the scarcity structure that existed before GenAI?
• What new competition between units or functions is appearing around the emerging constraint?
• Should the organization increase the scarce resource, redesign the workflow around it, or deliberately limit the demand placed upon it?

Review Capacity Becoming Scarce as GenAI Output Expands

• How much GenAI-generated or GenAI-assisted material now reaches human reviewers?
• Which outputs genuinely require expert verification and which can tolerate lighter assurance?
• Is review demand growing faster than production effort is falling?
• Where are scarce specialists becoming the limiting capacity in the workflow?
• Could sampling, exception review, or risk-tiered verification replace uniform line-by-line review?
• At what point does the cost of reviewing additional GenAI output exceed the value of producing it?
• What quality or error signals would show that review capacity has been exceeded?

Allocating Scarce Command and Expert Judgment Across GenAI-Expanded Work

• Which decisions genuinely require command judgment rather than routine processing or delegated authority?
• Where is specialist judgment most valuable because an error would have significant military consequences?
• Is GenAI increasing the number of issues presented for command or expert attention?
• Which judgments could move to lower levels or be handled through predefined rules without weakening accountability?
• Where could experts supervise exceptions and system performance instead of examining every individual output?
• What low-value work should stop consuming scarce judgment simply because GenAI can now generate it cheaply?
• How should judgment be allocated when the volume of machine-assisted work grows faster than the supply of experienced decision-makers?

Reassessing Reporting When Summarization Becomes Cheap

• Why does each recurring report exist: information production, accountability, coordination, record keeping, or command visibility?
• Which reports mainly compensate for the historical difficulty of consolidating and interpreting information?
• Could some recurring products be replaced by reliable on-demand access to authoritative underlying information?
• How should generated narrative remain distinguishable from the data or evidence on which it is based?
• Is cheap summarization encouraging more upward reporting than leaders actually need?
• Which reporting requirements should remain because the reporting act itself performs a necessary control or synchronization function?
• What reporting structure would minimize production and consumption burden while preserving military usefulness?

Staff Work Expanding Because GenAI Makes Cognitive Production Cheap

• Which staff activities are expanding primarily because GenAI has lowered the effort required to produce them?
• Does the additional analysis, planning, reporting, briefing, or documentation create proportional military value?
• What downstream review, coordination, meeting, or leadership-attention burden does each additional product generate?
• Are productivity gains being consumed by new expectations for more polished, more frequent, or more numerous outputs?
• Which forms of staff work should remain deliberately scarce even when production cost is low?
• What stopping rule would prevent low-value cognitive demand from expanding indefinitely?
• How would we know that GenAI is making the headquarters busier without making it more militarily effective?

Managing Convergence in GenAI-Assisted Staff Analysis

• Are supposedly independent analyses drawing on the same model, retrieval corpus, generated framing, or assumptions?
• Does shared GenAI improve necessary common understanding or reduce genuinely independent interpretation?
• Which conclusions might appear strongly corroborated even though they share a common machine-generated origin?
• How can the staff detect correlated analytical error before apparent consensus becomes misleading?
• Which questions warrant independent initial judgment, alternative sources, different methods, or deliberate red-team analysis?
• Where is convergence useful for routine coordination and where is cognitive diversity especially valuable?
• What provenance should be retained so decision-makers can tell whether agreement reflects independent reasoning or a shared GenAI influence?

Recognizing Organizational Structures Built Around Scarce Cognitive Capacity

• Which staff roles or organizational layers exist mainly to collect, aggregate, translate, summarize, draft, or move information?
• Which of those functions become materially less labor-intensive as GenAI improves?
• What other functions do the same roles provide through judgment, trust, relationships, authority, coordination, or institutional memory?
• Could some layers shrink or change purpose without weakening those nonproduction functions?
• What new roles become more important around data, evaluation, integration, assurance, dependency management, or human-machine workflow design?
• How would structural changes affect resilience if GenAI support were degraded or unavailable?
• What evidence would justify redesigning the organization rather than simply adding GenAI to the existing structure?

Redefining Expertise, Staffs, and Command

Assessing the Shift From Production Expertise Toward Evaluation Expertise

• Which parts of this expert role are becoming easier because GenAI can produce a credible first answer?
• Which parts still require human framing, contextual judgment, integration, challenge, or responsibility?
• What expertise is needed to recognize an answer that is plausible but wrong in ways a nonexpert would miss?
• Is the expert becoming a reviewer of more machine-assisted work, and can review capacity keep pace with that volume?
• Which production skills still need to be retained because performing the work contributes to the ability to evaluate it?
• How should expert development change when routine production provides fewer opportunities to build judgment through practice?
• What would show that GenAI has changed the nature of expertise rather than simply making experts more productive?

Preserving Expertise Needed to Challenge GenAI

• What must personnel know independently before they can meaningfully challenge GenAI output in this role?
• Which errors would be difficult to detect if users no longer practiced the underlying task themselves?
• What knowledge, methods, or professional instincts are most vulnerable to erosion through routine cognitive offloading?
• Which roles need enough manual competence to reconstruct the work when GenAI is wrong, unavailable, or outside its validated conditions?
• How can essential skills be practiced often enough to remain usable without duplicating every GenAI-supported task?
• What exercises would reveal whether personnel can still recognize and explain why a convincing GenAI answer is wrong?
• At what point would loss of underlying human expertise make nominal human oversight ineffective?

Assessing How GenAI Changes the Functions of Military Staffs

• Which staff activities primarily involve collecting, processing, summarizing, drafting, or transmitting information that GenAI can increasingly support?
• Which staff activities become more important as cognitive production becomes abundant?
• How could staff effort shift toward framing, integration, prioritization, verification, challenge, and coordination?
• Which boundaries between staff functions become less useful when information can be processed across specialties more easily?
• What new work appears around data, evaluation, assurance, integration, model management, or human-machine workflow design?
• How should the division of labor between generalists, specialists, technical personnel, and commanders change?
• What would a staff designed around abundant machine-assisted cognition do differently from a staff that simply adds GenAI to existing routines?

Reconsidering Headquarters Size and Composition

• How much headquarters workload exists because cognitive production currently requires large amounts of human labor?
• Which of that work could GenAI reduce without weakening judgment, coordination, relationships, authority, or accountability?
• Would released capacity actually permit a smaller headquarters, or would the organization use it to produce more work?
• Which specialist roles could become more important even if routine production staffing declines?
• What new requirements for data, evaluation, integration, assurance, resilience, and technical support would offset potential reductions elsewhere?
• How would a smaller GenAI-dependent headquarters perform if its preferred systems became degraded or unavailable?
• What evidence would justify changing headquarters size or composition rather than simply changing how existing personnel work?

Evaluating Whether GenAI Should Expand Span of Control

• Which information-processing burdens currently limit the number of subordinate elements a commander can supervise effectively?
• Does GenAI reduce those burdens enough to change the commander's real capacity for understanding and interaction?
• Would a wider span preserve meaningful relationships with subordinate leaders, or only increase informational visibility?
• Can subordinate commanders retain adequate decision authority if higher headquarters can monitor more activity at once?
• Does additional information create more supervisory work that could offset the apparent increase in command capacity?
• How would the wider span perform if GenAI support or connectivity were degraded?
• What evidence would show that a wider span improves command effectiveness rather than simply making broader supervision technically possible?

Managing GenAI-Enabled Micromanagement

• What subordinate activity has become newly visible or easier for higher headquarters to interpret because of GenAI?
• Are senior leaders intervening more often because information is available, even when subordinate judgment remains sufficient?
• Does greater visibility give higher headquarters genuine contextual understanding or only the appearance of it?
• Which decisions should remain at lower levels even when higher headquarters can observe them in detail?
• Are reporting, monitoring, or AI-generated recommendations unintentionally encouraging more centralized intervention?
• How can commanders distinguish information they should know from information they should personally act upon?
• What command practices would preserve subordinate initiative while still exploiting the improved visibility GenAI provides?

Reassessing How Authority Is Distributed Across Command Levels

• How has GenAI changed access to useful analysis and information at different levels of command?
• Which decisions could move downward because subordinate leaders now have better analytical support?
• Which decisions could benefit from higher-level integration because GenAI improves visibility across multiple subordinate elements?
• Where does local context still outweigh the broader information available at higher headquarters?
• Which decisions require centralization for coherence, scarce-resource allocation, legal authority, or cross-unit coordination?
• How would a change in authority distribution affect responsiveness, accountability, initiative, and resilience?
• What decision-by-decision evidence would justify moving authority rather than assuming GenAI should generally centralize or decentralize command?

Maintaining Command Responsibility in GenAI-Supported Decisions

• Who is formally responsible when GenAI materially shapes the analysis, options, or recommendation behind a command decision?
• Does the responsible commander understand what role GenAI played and which parts of the advice depend on machine-generated reasoning?
• Can the commander identify the key evidence, assumptions, uncertainties, and limitations behind the recommendation?
• Is there enough time and access to underlying information for the commander to challenge the recommendation meaningfully?
• Could human approval become a procedural signature even when the human lacks the competence or information needed for genuine judgment?
• Which decisions require stronger evidence of human understanding because their consequences are particularly serious or difficult to reverse?
• What conditions must exist for human responsibility to remain substantive rather than merely formal?

Calibrating Command Reliance on GenAI Decision Support

• For which kinds of decisions has the GenAI support demonstrated enough reliability to influence command judgment?
• How should reliance change with task, consequence, uncertainty, data quality, model version, and operating environment?
• What warning signs should cause a commander to reduce or suspend reliance on the system?
• Where should GenAI remain one analytical input rather than become the primary framing of the problem?
• How do commanders learn the characteristic situations in which the system performs well and those in which it does not?
• What independent information or judgment remains available when the GenAI recommendation is uncertain or contested?
• How can training produce calibrated reliance rather than either habitual acceptance or habitual distrust?

Preserving Command Judgment Through Routine GenAI Use

• Is the commander still forming an independent understanding of important problems before seeing a GenAI-generated framing or recommendation?
• Which elements of command judgment could weaken if GenAI routinely performs the initial analysis and option development?
• Are leaders still engaging directly enough with evidence, subordinate perspectives, uncertainty, and operational context?
• Could repeated exposure to fluent recommendations anchor commanders before alternative framings are considered?
• Which decisions or exercises should require independent human assessment before GenAI support is introduced?
• How well could commanders continue framing problems and making decisions if the GenAI layer disappeared unexpectedly?
• What pattern of routine use would strengthen command judgment rather than gradually make it dependent on machine preparation?

Managing GenAI-Driven Capability Evolution

Reassessing a Fielded Capability After a Major GenAI Capability Shift

• Which assumptions in the current capability design were based on limitations that the new GenAI capability may have removed?
• What military task or constraint has become technically different enough to justify reassessment?
• Does the new capability change the military outcome that can be achieved, or mainly the efficiency of the existing approach?
• Which training, data, infrastructure, assurance, staffing, doctrine, or authority assumptions would change if the new functionality were adopted?
• What new dependencies or failure modes would the change introduce into an otherwise stable capability?
• Which parts of the existing capability remain valid and should be preserved?
• What operational evidence is needed before changing a fielded capability in response to the new GenAI possibility?

Deciding Whether to Retain, Integrate, Adapt, or Redesign an Existing System

• Does the existing system still meet its military requirement well enough that no GenAI-driven change is necessary?
• Can useful GenAI functionality be added without altering the basic architecture, workflow, or operating concept?
• Which limitations require broader adaptation of data, interfaces, staffing, training, or procedures rather than simple integration?
• Has incremental modification become so complex that fundamental redesign would provide greater long-term value?
• What reliability, interoperability, resilience, or institutional knowledge would be lost by replacing the current system?
• How do the options differ in cost, fielding time, flexibility, switching risk, and ability to absorb future GenAI change?
• Which response solves the military problem without assuming that newer GenAI automatically justifies deeper redesign?

Managing Requirements That Change Faster Than Acquisition Cycles

• Which requirements express enduring military outcomes and which are tied too closely to today's GenAI implementation?
• What technical assumptions are likely to become obsolete before the acquisition process delivers the capability?
• Could requirements specify performance, operating conditions, interoperability, and unacceptable failure rather than a particular model or provider?
• Which implementation decisions can remain open until later without threatening the broader program?
• How can user feedback and evolving technical evidence change lower-level requirements without repeatedly reopening the military need?
• What testing and assurance requirements must remain stable even when the underlying models change?
• How should the acquisition process preserve enough flexibility to exploit improvement without turning every new model release into a new requirement?

Reassessing Doctrine After GenAI Changes Available Options

• Which doctrinal assumptions depend on previous limits in staff capacity, information processing, software production, or decision support?
• What genuinely new military option has GenAI made feasible rather than merely easier to describe?
• Which physical, institutional, legal, or organizational constraints still limit the apparent new option?
• Has the new practice survived enough exercises or operational use to justify doctrinal attention?
• Should the lesson remain a local procedure or technique until its underlying mechanism is better understood?
• Which doctrinal principles remain valid even though the technical means of applying them are changing?
• What evidence would justify changing doctrine without encoding temporary characteristics of today's GenAI systems?

Assessing When Training Baselines Become Obsolete Faster Than Training Cycles

• Which trained procedures depend on model behavior, interfaces, limitations, or functionality that has materially changed?
• Are personnel being taught workarounds for limitations that no longer exist or missing new failure modes introduced by newer systems?
• Which competencies should remain stable even when the specific GenAI tool changes?
• Which current implementation details need a rapidly updateable training layer rather than a complete curriculum rewrite?
• Can operational units receive credible updates faster through short modules, exercises, or local instruction?
• Could constant interface and capability change prevent personnel from developing deep competence with the underlying military task?
• How should qualification remain meaningful when different units may be operating different versions of the capability?

Managing Continuous Model and Software Change in Fielded Capability

• Which model, software, data, integration, provider, or configuration changes can materially alter the fielded capability's behavior?
• How is the exact operational configuration recorded so users and evaluators know what is actually fielded?
• Which changes can move through routine release processes and which require additional operational approval?
• Can updates be staged, rolled back, or temporarily frozen during sensitive operational periods?
• How are users informed when a change alters known limitations, workflow assumptions, or expected performance?
• What external provider changes can affect the capability even when the military organization releases nothing itself?
• What configuration-management process allows continuous improvement without losing control of the operational baseline?

Retesting Capabilities After Model or Data Changes

• What exactly changed in the model, data, retrieval system, configuration, software, provider, or operating context?
• Which previously validated performance claims depended on something affected by that change?
• What mission-relevant tests need to be repeated before users rely on the updated capability?
• Could an apparent improvement in one area create a new failure in another task, population, language, or operating condition?
• When is focused regression testing sufficient and when does the change justify broader revalidation?
• Does the update alter human workload, reliance, security, degraded-mode behavior, or interoperability even if headline model performance improves?
• What evidence is sufficient to restore confidence without making continuous updates impossible to field?

Coordinating Doctrine, Training, Systems, and Organization as One Changes

• What other elements of the military capability are affected when this GenAI component changes?
• Does the technical change require different training, procedures, staffing, authorities, doctrine, infrastructure, or evaluation?
• Which capability elements are changing faster than the others can absorb?
• What operational problems appear when users, systems, doctrine, and assurance are based on different capability assumptions?
• Who is responsible for maintaining visibility across the whole capability rather than only the technical component?
• Which changes must be synchronized before the new GenAI functionality can create military value?
• How should the integrated capability baseline be maintained as individual elements continue to evolve?

Managing Diverging GenAI Capability Baselines Across the Force

• Which units or functions are operating with materially different GenAI models, configurations, data access, or functionality?
• Do those differences affect interoperability, training, support, expectations, or command planning?
• Which variation represents useful local experimentation and which is producing unnecessary fragmentation?
• Can personnel move between units without having to relearn fundamentally different ways of performing the same task?
• Do commanders know which GenAI capabilities and limitations actually exist across the formations they control?
• What common minimum baseline is necessary for shared operations while still allowing local innovation?
• How can successful local evolution spread across the force without forcing every unit onto an identical technical implementation?

Assessing When Existing Systems or Processes Become Capability Constraints

• What military capability is being limited by the existing system or process?
• Is the constraint caused by closed architecture, poor data access, integration difficulty, security rules, authority, workflow design, or something else?
• Could the limitation be mitigated while preserving the useful parts of the existing arrangement?
• Does the constraint materially affect readiness, learning, tempo, resilience, interoperability, or ability to exploit new GenAI capability?
• Are we treating the system as obsolete because it is old, or because it genuinely prevents a required military outcome?
• How much future adaptation effort would be spent working around the constraint if it remains?
• At what point do the accumulated costs and lost options make redesign more sensible than continued adaptation around it?

Managing GenAI Ecosystem Dependence and Military Autonomy

Relying on Commercial GenAI Providers for Military Work

• Which military functions now depend on a commercial model, platform, API, tool, or service?
• What aspects of availability, pricing, updates, technical behavior, support, and permitted use remain under provider control?
• Could the provider restrict, alter, or withdraw the capability during a crisis or conflict?
• What military data, security, assurance, or jurisdictional risks accompany the arrangement?
• How quickly could the affected function move to an alternative if the provider became unavailable?
• What capability, innovation speed, scale, or expertise makes the commercial dependence worthwhile?
• Which commercial dependencies remain acceptable because loss is tolerable, and which could become constraints on military freedom of action?

Assessing Dependence on Foreign Models or Infrastructure

• Which important GenAI capabilities depend on technology, infrastructure, or expertise controlled under another jurisdiction?
• What export controls, sanctions, licensing decisions, legal authorities, or political changes could affect continued access?
• Where are the relevant models, data, cloud services, hardware, support personnel, and supply-chain components actually located?
• Would foreign control create different risks in peacetime, crisis, and conflict?
• What domestic or trusted-partner alternatives could replace the dependency, and at what loss of capability?
• How long would reducing the dependence take if geopolitical conditions changed suddenly?
• What level of foreign dependence is compatible with the military function's required freedom of action?

Evaluating Cloud and Compute Dependence

• Which GenAI functions require continuous access to external cloud or high-performance compute?
• What happens to those functions if connectivity, power, capacity, authentication, or provider access is degraded?
• Which workloads genuinely require centralized frontier-scale compute and which could run on smaller local systems?
• What performance reduction would be acceptable if the force had to shift from cloud to edge or local inference?
• Are multiple critical functions concentrated on the same cloud, region, identity service, or compute infrastructure?
• How much compute capacity must remain assured during crisis or conflict?
• What architecture best balances access to high-performance compute with the need to continue operating under denial or disconnection?

Managing Semiconductor and Hardware Supply Dependence

• Which GenAI capabilities depend on specialized chips, memory, networking equipment, or other hardware with concentrated supply?
• Where are the critical components designed, fabricated, packaged, assembled, and tested?
• Which geographic, industrial, export-control, or geopolitical disruptions could constrain replacement or expansion?
• How much inventory, substitution capacity, or alternative hardware is available if normal supply is interrupted?
• Could more efficient models or alternative architectures reduce dependence on the most constrained hardware?
• How long could the force sustain and expand the capability during a serious supply disruption?
• Which hardware dependencies are consequential enough to require strategic mitigation rather than ordinary procurement management?

Responding When Commercial GenAI Evolves Faster Than Military Absorption

• Which commercially available GenAI capabilities are changing faster than the military can evaluate, accredit, integrate, train, and procure?
• Are current processes delivering technology only after the assumptions behind the original requirement have changed?
• Which security, reliability, legal, and operational controls remain necessary despite pressure to move faster?
• Where could reusable infrastructure, modularity, preapproved environments, or rapid experimentation reduce absorption time?
• How can the military evaluate capability families or mission effects without restarting the entire process for every model release?
• What military opportunity is being lost because the institution adapts more slowly than the commercial technology changes?
• How can the organization remain responsive without allowing commercial release cycles to dictate military priorities?

Preparing for Provider Policy, Pricing, or Access Changes

• Which military functions depend on provider terms that can change without military approval?
• What pricing, licensing, geographic, acceptable-use, API, feature, or support changes would materially affect the capability?
• How much warning is the military likely to receive before a consequential provider change takes effect?
• What contractual rights exist for notification, continued service, data extraction, portability, and transition support?
• Which technical alternatives could be activated if the new terms became unacceptable?
• How much switching cost has accumulated through data, workflows, integrations, training, or dependent applications?
• What contingency is proportionate to the military consequence of losing acceptable access to the provider?

Deciding Which GenAI Capabilities Require Assured Access

• Which GenAI-supported functions would create a material military capability gap if access disappeared?
• How long can each function tolerate loss before performance becomes unacceptable?
• Could the function continue through a lower-capability model, local system, alternative provider, or human workflow?
• What level of contractual, technical, geographic, dedicated, redundant, or sovereign assurance matches the consequence of failure?
• Does the cost of stronger assurance exceed the military value of guaranteeing the capability?
• Which uses should remain opportunistic precisely because their loss would be tolerable?
• What evidence would justify moving a GenAI function from useful support into the category of capability that must be assured?

Balancing Commercial Innovation With Military Autonomy

• What military advantage comes from using the fastest-moving commercial GenAI ecosystem?
• Which forms of decision, data, technical, operational, legal, or evolutionary autonomy would be reduced by deeper dependence?
• Where would accepting a lower-performing controlled capability be preferable to relying on a superior but externally constrained one?
• Could a mixed architecture combine commercial frontier capability with controlled fallback or sovereign components?
• Which autonomy requirements become more important specifically during crisis, conflict, or external restriction?
• How much innovation or performance would be sacrificed by demanding greater control?
• What balance preserves enough freedom of action without isolating the force from valuable commercial advances?

Assessing Switching Costs Between GenAI Providers

• What would actually have to change if the military moved this function to another provider?
• Which data structures, APIs, retrieval systems, fine-tuning, workflows, evaluations, interfaces, or cloud services are provider-specific?
• How much user training and operational procedure is tied to the current system's behavior?
• Can existing mission evaluation suites be applied directly to a replacement provider?
• How long would a switch take under normal conditions and under an urgent loss of service?
• Which parts of the capability are technically portable but operationally difficult to migrate because of approval, training, or integration requirements?
• What portability should be preserved now to keep future switching from becoming prohibitively disruptive?

Identifying Hidden Dependencies Across the GenAI Ecosystem

• What models, cloud services, APIs, data feeds, identity systems, libraries, hardware, networks, power, and specialists sit behind the visible capability?
• Which of those dependencies are controlled by organizations outside the primary contractual relationship?
• Are supposedly independent GenAI capabilities relying on the same hidden upstream provider or infrastructure?
• Which dependency would take longest to replace if it failed?
• Are critical functions dependent on a small number of technical personnel whose knowledge is difficult to reproduce?
• Could one shared failure disable several apparently separate GenAI-enabled functions at once?
• What does a full dependency map reveal that an ordinary list of direct suppliers would miss?

Balancing Collective GenAI Capability and Common-Mode Risk

Operating With Uneven GenAI Capability Across Partners

• How different are partners in model access, data, infrastructure, integration, expertise, evaluation, and operational reliance on GenAI?
• Which of those differences materially affect a shared military process rather than remaining internal national variation?
• Is a more mature partner producing information or decisions at a rate others cannot absorb?
• Where could a less mature partner become the limiting stage in collective planning, coordination, or information exchange?
• Can stronger partners provide services or support without making the coalition dependent on capabilities only they can supply?
• Which differences need common standards and which can safely remain national?
• How is uneven GenAI maturity changing the practical distribution of roles and influence within the collective force?

Sharing GenAI Services Across Allied Organizations

• Which GenAI capabilities create enough common value to justify shared provision rather than separate national solutions?
• Would sharing models, infrastructure, data services, evaluation, or expertise improve access and reduce duplication?
• Who controls the shared service and decides its models, priorities, operating rules, and acceptable uses?
• How are national caveats, classification, data sovereignty, security, cost, and capacity allocation handled?
• What happens if a participating nation, provider, or shared infrastructure component becomes unavailable?
• Does the shared service create a common dependency whose failure would affect multiple partners at once?
• Which functions should use the shared capability while preserving an independent national fallback?

Assessing Whether National GenAI Advantage Transfers to Collective Operations

• Which part of the national GenAI capability can actually be shared with partners during operations?
• Can the outputs be released at the classification, speed, and level of detail the coalition requires?
• Do partners have enough access to underlying evidence to understand and trust the AI-assisted product?
• Can their technical systems and staff processes absorb the faster or richer national output?
• Which national advantages depend on proprietary data, infrastructure, authorities, expertise, or integration that cannot transfer?
• Does reliance on the strongest national capability create a new collective single point of failure?
• What evidence would show that the national advantage is improving collective military performance rather than only national performance?

Maintaining Collective Tempo Across Different GenAI Maturity Levels

• Which shared military activities require partners to synchronize despite different internal GenAI-enabled speeds?
• Where can partners operate asynchronously without reducing collective effectiveness?
• Is the fastest partner creating more information, requests, or changes than others can process?
• Which collective process is currently governed by the slowest necessary participant?
• Could shared services, hosted access, translation, or procedural changes reduce the practical tempo gap?
• What minimum product, response time, or interface does every participant need to meet for the collective process to work?
• How can the coalition accommodate speed asymmetry without forcing either universal adoption of the leading technology or universal operation at the slowest internal tempo?

Balancing Common GenAI Platforms With National Flexibility

• Which interoperability problems would a common GenAI platform solve?
• Which national missions, laws, sovereignty requirements, security rules, or innovation needs make complete standardization undesirable?
• Would a common platform concentrate too much dependence on one provider, architecture, model family, or governance regime?
• Which interfaces, data standards, metadata, security controls, and assurance expectations actually need to be common?
• What implementation choices can remain national without undermining collective operations?
• How easily could one nation change provider or architecture while remaining interoperable with the group?
• What degree of commonality gives the coalition reliable interoperability without sacrificing useful national flexibility and resilience?

Assessing Whether Data Asymmetries Limit Collective GenAI Capability

• Which partners have materially different access to the data required for the shared GenAI-supported function?
• Are those differences caused by classification, releasability, national caveats, law, technical access, language, quality, or data governance?
• Do partners using the same model receive different capability because their usable information environments differ?
• Which collective processes are constrained by the participant with the least access to necessary data?
• Can metadata, data-centric security, or different release mechanisms improve sharing without removing legitimate controls?
• Are language or data-quality differences creating unequal model performance across partners?
• How much of the collective capability gap comes from data asymmetry rather than differences in GenAI technology?

Assessing Common-Model Dependence Across Partners

• How many partners depend on the same model family, provider, or model-serving infrastructure for consequential military work?
• What collective functions would be affected if that model failed, changed behavior, or became unavailable?
• Could one model weakness or adversarial technique create similar errors across several national organizations?
• Are partners evaluating the common model independently or largely relying on shared assumptions about its performance?
• What genuinely independent alternatives remain available if confidence in the common model collapses?
• Does common model use create enough interoperability and shared learning to justify the concentration risk?
• What level of model diversity would reduce common-mode failure without creating unnecessary fragmentation?

Assessing Cognitive Diversity With Shared GenAI Systems

• Are national staffs reaching more similar conclusions because they begin from the same models, sources, or generated summaries?
• Does that convergence reflect useful common understanding or reduced independence of reasoning?
• Which shared assumptions or framing errors could propagate across several partners without being recognized as common-origin effects?
• Are national doctrine, experience, culture, and specialist knowledge still producing genuinely distinct judgments?
• How could decision-makers tell whether apparent multinational agreement reflects independent corroboration?
• Which consequential questions warrant separate sources, methods, models, or independent first judgments?
• Where does preserving cognitive diversity add resilience, and where would it merely create unnecessary analytical inconsistency?

Evaluating Correlated Failure Across Shared GenAI Infrastructure

• Which partners rely on the same cloud, identity service, data platform, model provider, network, chip supply, or software component?
• Which single upstream failure could disable several national capabilities simultaneously?
• Are supposedly redundant national alternatives actually dependent on the same hidden infrastructure?
• Could one cyber compromise, provider outage, update failure, or supply disruption propagate across the collective force?
• Which independent operating paths would remain after the common dependency failed?
• How much integration complexity would additional infrastructure diversity create?
• What combination of shared infrastructure and independent fallback provides useful resilience without destroying interoperability?

Responding When One Partner's GenAI Dependency Constrains the Group

• Which collective military function depends on GenAI capability supplied primarily by one partner?
• What happens to collective tempo, information flow, or decision quality if that partner loses the capability?
• Can another participant substitute for the lost function without major technical or procedural change?
• Can the affected partner revert to a lower-capability mode while remaining interoperable with the coalition?
• Has the group become dependent on an external provider or infrastructure indirectly through that partner?
• Which critical functions should be redistributed before the dependency becomes operationally intolerable?
• How should collective plans account for partners degrading asymmetrically rather than assuming everyone loses or retains GenAI capability together?

Maintaining Military Capability Under GenAI Degradation

Operating After Loss of a Mission-Critical GenAI Service

• Which military functions stop, slow, or degrade when the GenAI service disappears?
• How long can each affected function tolerate the loss before mission performance becomes unacceptable?
• What minimum military capability must continue even without the preferred service?
• Which activities can be delayed or dropped so scarce human or fallback capacity can support the most important functions?
• What alternative systems, procedures, personnel, or data are immediately available?
• How long does it take to move from normal operation to a viable degraded mode?
• What does the outage reveal about dependencies that had become more consequential than leaders realized?

Working Through Degraded Connectivity to GenAI

• Which GenAI-supported functions assume continuous connectivity to remote models, data, identity services, or compute?
• What capability remains under intermittent, low-bandwidth, high-latency, or disconnected conditions?
• Which models, data, reference material, and retrieval resources should be available locally?
• Which tasks can tolerate delayed synchronization and which require immediate local processing?
• What should shift to manual methods when remote GenAI access becomes unreliable?
• How should scarce bandwidth or compute be prioritized among competing GenAI-supported functions?
• What architecture would allow the military function to degrade gradually rather than fail when connectivity disappears?

Responding to Compromised or Unreliable GenAI Outputs

• What signs indicate that the problem exceeds ordinary model error and may involve corrupted data, manipulation, configuration failure, or compromise?
• Which users, decisions, and products may already have relied on the suspect outputs?
• Can the affected model, data source, retrieval corpus, tool, or integration be isolated without disabling unrelated functions?
• What independent evidence or known-good process can be used to reassess consequential outputs?
• When should reliance be reduced or suspended even before the precise cause of the problem is known?
• How will users know when restored output is trustworthy enough for operational reliance to resume?
• What military capability must remain available while the organization investigates and rebuilds confidence?

Switching to Alternative GenAI Support Under Constraint

• Is the fallback GenAI capability actually available under the same conditions that disabled the primary system?
• Does the alternative depend on different cloud, network, identity, data, hardware, and provider infrastructure?
• What performance, data access, functionality, or security will be lost after the switch?
• Can existing workflows and applications use the alternative without extensive redesign?
• Are personnel sufficiently familiar with the fallback to use it effectively under time pressure?
• Has the replacement been tested against the mission rather than merely confirmed to be technically accessible?
• How long would it take to reach a viable military operating level after the switch begins?

Falling Back to Human-Only Workflows

• Which essential activities can personnel still perform without any GenAI support?
• Do the people assigned to the fallback retain the underlying knowledge and practical skill to perform the work independently?
• What procedures, reference materials, data, authorities, and non-AI tools are available for the human-only mode?
• How much slower or less capable will the function become once human capacity is again the limiting resource?
• Which tasks must be stopped or deprioritized because the full AI-enabled workload cannot be reproduced manually?
• How long can the reduced human-only mode be sustained with available staffing?
• What minimum manual capability is worth preserving even if it appears inefficient during normal GenAI-supported operations?

Maintaining Critical Functions With Reduced GenAI Support

• Which military functions require the highest remaining GenAI capability during degradation?
• Which tasks can tolerate smaller models, stale data, increased latency, fewer features, or reduced automation?
• What minimum machine-assisted capability preserves a meaningful military advantage over full manual fallback?
• How should scarce compute, bandwidth, model access, or technical support be allocated among competing users?
• Which lower-priority activities should revert to manual methods so critical functions retain machine support?
• What changes in speed, quality, confidence, or risk should commanders expect in the reduced mode?
• How should degraded GenAI performance be incorporated into realistic readiness standards and operational planning?

Preserving Human Competence for Degraded Operations

• Which human skills become essential when GenAI support is unavailable, unreliable, or suspected of compromise?
• Are those skills practiced often enough to remain usable under operational pressure?
• Which competencies are decaying because personnel rarely perform the underlying work without GenAI?
• Which critical roles need deeper independent capability than ordinary users?
• How can the organization preserve necessary human competence without duplicating all routine GenAI-supported effort?
• What AI-denied drills or exercises would reveal whether personnel can actually continue the mission?
• Which human skills should be treated as resilience assets even if normal operations rarely require them?

Testing Mission Capability Without Preferred GenAI Support

• Which consequential GenAI dependency should be deliberately removed, degraded, or made untrustworthy during the test?
• What military functions must participants still perform despite the disruption?
• Do personnel recognize the failure and switch to the correct degraded mode without being prompted?
• Does the designated alternative actually work with the required data, permissions, interfaces, and staffing?
• What level of performance degradation remains acceptable and what indicates a readiness failure?
• Which previously unknown technical, organizational, or human dependencies become visible during the exercise?
• Does the test include recovery and revalidation, or does it stop once the fallback has been demonstrated?

Assessing GenAI Dependence Before a Failure

• Which military functions would be materially affected if their preferred GenAI support disappeared tomorrow?
• How quickly would performance decline from inconvenience to unacceptable mission degradation?
• What human, technical, procedural, and data fallback capability actually remains?
• How long would it take to move each critical function into a viable alternative mode?
• Do supposedly redundant alternatives share the same hidden infrastructure or provider dependencies?
• Is normal-mode performance improving while the degraded-mode capability floor is quietly falling?
• What level of GenAI dependence is acceptable given the military consequence, fallback quality, and time required for recovery?

Recovering From a Shared GenAI Ecosystem Outage

• Which military functions, units, partners, services, and providers are affected by the common outage?
• What shared dependency caused capabilities that appeared separate to fail together?
• Which military functions should be restored first based on mission consequence and tolerated downtime?
• What independent alternatives can reduce demand on the disrupted ecosystem during recovery?
• Are authentication, documentation, recovery tools, specialists, or fallback systems themselves dependent on the failed environment?
• What technical and mission-relevant validation is required before restored services are trusted again?
• What dependency exposed by the outage should be redesigned rather than simply restored to its previous state?

Anticipating Strategic Surprise From GenAI-Enabled Adaptation

Detecting Adversary Capability Gains Without New Platforms

• Which military functions could improve substantially without any visible change in platforms, force structure, or major procurement?
• Are there signs of faster analysis, planning, software development, information exploitation, coordination, or organizational learning?
• What changes in operational behavior would be expected if the apparent cognitive improvement were becoming militarily consequential?
• Which digital, workforce, data, or organizational developments could explain the gain despite weak traditional modernization signatures?
• Could the observed improvement result from better doctrine, communications, delegation, training, or another factor rather than GenAI?
• Is the change recurring across units and missions, or limited to an unusually capable team or circumstance?
• What combination of nonmaterial indicators would justify revising our assessment of the adversary's military capability?

Reassessing Intelligence Indicators for GenAI-Enabled Change

• Which current indicators assume that meaningful military modernization will produce visible procurement, platform, infrastructure, or force-structure changes?
• What organizational, digital, data, workforce, commercial, training, and behavioral indicators could reveal GenAI adaptation earlier?
• Which apparent indicators are too common during experimentation to provide strong evidence on their own?
• What pattern would show movement from access and experimentation into recurring military integration?
• Which indicators should be tracked longitudinally so changes in adaptation rate become visible?
• How can multiple independent indicator types be combined without treating correlation as proof of GenAI causation?
• What important capability change could existing collection and assessment methods miss because it occurs inside ordinary organizational processes?

Recognizing Military Effects From Commercial GenAI Diffusion

• Which commercially available GenAI capabilities could remove a meaningful military limitation without requiring a major defense-development program?
• Which actors can acquire those capabilities through civilian markets, open models, cloud services, or commercial software?
• What data, expertise, infrastructure, integration, authority, or operational access is still required before commercial capability becomes militarily useful?
• Which specialist cognitive advantages become easier to approximate as advanced GenAI diffuses?
• How quickly could a commercially available capability move from acquisition into consequential military use?
• Which physical, organizational, security, or institutional constraints would continue to limit the military effect?
• What indicators would show that commercial diffusion has crossed from technological availability into sustained military capability?

Assessing Sudden GenAI Uplift in Smaller Actors or Proxies

• Which existing capability shortage is GenAI reducing most sharply for the smaller actor or proxy?
• Is GenAI substituting for scarce staff capacity, language skills, software expertise, research, analysis, planning support, or another cognitive resource?
• What military or security activity becomes feasible, cheaper, faster, or more scalable as a result?
• Which major disadvantages remain because they depend on physical assets, logistics, industry, sensors, trained forces, or secure infrastructure?
• Does external support from a sponsor, partner, cloud provider, or commercial ecosystem amplify the GenAI-enabled gain?
• Is the uplift narrow and task-specific, or is it spreading across several functions through organizational learning?
• At what point would a disproportionate cognitive gain materially change how the actor should be assessed despite continued weakness in overall military power?

Responding to Unexpected Changes in Adversary Decision Tempo

• What evidence shows that the adversary's consequential decision cycle has actually shortened?
• Which stage appears to have accelerated: sensing, analysis, option development, approval, dissemination, execution, or learning?
• Could GenAI plausibly explain the observed change, or do delegation, communications, doctrine, risk tolerance, or other reforms provide a stronger explanation?
• Is faster activity producing better operational effect, or merely more frequent decisions and outputs?
• Which of our own processes become disadvantaged by the new tempo difference?
• Could we respond by changing authority, information flow, coordination, or other bottlenecks rather than simply trying to make every process faster?
• What evidence would show that the tempo change is durable and institutionally supported rather than situational?

Recognizing When Institutional Adaptation Outpaces Existing Assumptions

• Which plans or assessments assume that the other force will adopt new capabilities at a historically slower rate?
• Are software, training, doctrine, acquisition, workflows, or command practices changing faster than those assumptions allow?
• Is operational feedback reaching developers, trainers, and decision-makers quickly enough to create repeated adaptation cycles?
• Are successful local practices spreading through the organization rather than remaining isolated?
• What evidence shows that the institution has become better at exploiting successive GenAI improvements rather than merely adopting one useful tool?
• Which of our own plans become fragile if the other actor can learn and reorganize faster than expected?
• How often should assumptions about the actor's adaptation rate be revisited when the underlying technology and institutional response are both changing rapidly?

Distinguishing Genuine Competitive Urgency From GenAI Arms-Race Rhetoric

• What specific military capability is claimed to be changing because of GenAI?
• What credible causal pathway connects the technological development to an operational or strategic effect?
• What evidence shows sustained exploitation rather than access, investment, demonstrations, procurement activity, or publicity?
• Is there a meaningful relative capability difference, or are both sides improving without a consequential change in military position?
• Are model benchmarks, spending, user counts, project numbers, or public statements being used as substitutes for evidence of military performance?
• What is the plausible cost of waiting if the concern is real, including lost learning and slower future adaptation?
• What practical response would improve underlying military capability without treating every GenAI advance as evidence of an arms race?

Identifying When a GenAI Adaptation Gap Becomes a Readiness Gap

• Which actual military requirement is becoming harder to meet because GenAI-related adaptation is lagging?
• Can the current force still meet the required standard through existing non-GenAI methods?
• Does the gap affect required speed, scale, quality, resilience, interoperability, learning, or another operationally relevant dimension?
• Is the shortfall caused by technology access or by weaker integration, evaluation, data, training, authority, or organizational learning?
• How quickly could the gap be closed if operational conditions made it immediately consequential?
• Has the difference progressed beyond modernization into an inability to perform an assigned mission to the required standard?
• What evidence would justify incorporating GenAI adaptation directly into formal readiness assessment?

Assessing the Significance of an Unexpected GenAI Capability Demonstration

• What exactly did the demonstration prove under the conditions that were actually shown?
• Which data, compute, human support, preparation, connectivity, or favorable assumptions contributed to the result?
• Can the capability perform reliably and repeatedly rather than succeed only in a carefully prepared demonstration?
• Can it operate at military scale, under realistic security constraints, and in degraded or adversarial conditions?
• What complementary infrastructure, data, expertise, authorities, training, and integration would be needed for operational use?
• Does the demonstration address a meaningful military constraint, and how difficult would it be to reproduce or field?
• What additional evidence would distinguish an isolated technical achievement from a narrow operational capability or deeper institutional exploitation competence?

Reassessing the Time Available to Adapt After GenAI Capability Shifts

• How quickly could the new GenAI capability diffuse to the actors for whom it would matter militarily?
• How long would our organization need to recognize the change, understand it, test it, decide on a response, acquire what is needed, train personnel, and integrate the result?
• Is our full adaptation cycle longer than the period during which the competitor's new capability could become consequential?
• Which parts of our response require accumulated experience that cannot be compressed simply by spending more money later?
• What preparations can be made now without knowing exactly which future model or application will matter?
• Which decisions become increasingly costly if delayed until the military effect is fully demonstrated?
• How should readiness, experimentation, modularity, evaluation capacity, and force development change if warning-to-effect timelines continue to shrink?