Why Space Operations Training Has a Structural Problem
The space domain has grown more contested, but training access has historically been the opposite: limited, scheduled, and disconnected from the actual operational environment.
The Space Force inherited an Air Force training architecture built around specific courses and exercise rotations. Legacy simulators ran scripted scenarios. Live orbital data—the source of truth for current threats—was available in operational systems but rarely integrated into training environments. The scenarios Guardians practiced were usually months or years behind the threats they would face, and access to any training environment beyond coursework required scheduling against limited simulator availability.
As adversary behavior in orbit became more sophisticated—non-nominal maneuvering, proximity operations, electronic warfare against space assets—the training fidelity gap widened. Space Training and Readiness Command (STARCOM), activated in August 2021 at Peterson Space Force Base, Colorado, inherited this architecture. Its core mission is individual training and career field development, but the quality of what it can produce is bounded by the quality of the environments it can use. OTTI is the Space Force's program for changing those environments. Its investment premise is that training infrastructure needs to be as live and adversarially realistic as the operational picture—not a years-lagged approximation of it.
The Two-Contract Sequence
The January 2026 award to Slingshot was $27 million via a Commercial Solutions Opening, an acquisition authority designed for commercial technology that doesn't fit existing procurement categories. The 18-month contract built on a $25 million Strategic Funding Increase (STRATFI) that Space Systems Command and SpaceWERX awarded Slingshot in 2022. Its specific focus was integrating AI-native scenario training with STARCOM's existing training enterprise. Slingshot says STARCOM tested its TALOS AI agent during this period as a potential training force multiplier—that characterization comes from Slingshot's own materials and should be held as vendor-attributed until STARCOM publishes an independent assessment.
The July 2026 MENTAT award was a SBIR Phase III contract—the non-competitive commercialization pathway available to companies with demonstrated DoD utility after completing earlier SBIR phases. At $69.2 million over 4.5 years, it is the largest award in Slingshot's history. MENTAT is designed for home-station access, letting Guardians rehearse protect-and-defend scenarios and test courses of action from their normal work environments rather than requiring access to specific simulation facilities.
The acquisition sequence—SpaceWERX STRATFI, then CSO, then SBIR Phase III—shows the Space Force using commercial pathways to develop and scale a training capability faster than traditional Major Defense Acquisition Program timelines would allow. That tradeoff is deliberate and often correct. It also concentrates the capability in a single commercial vendor and embeds a technology architecture before the operational community has fully characterized what the system needs to do across the full range of training requirements.
What TALOS and MENTAT Are, and What Remains Vendor Claims
Slingshot describes TALOS as an AI-powered operational training and strategy agent. Published descriptions say it models spacecraft behaviors, generates strategic response options, processes large volumes of orbital data, and supports scenario rehearsal across variable conditions. MENTAT is the mission rehearsal application layer on top of TALOS—the interface through which Guardians access scenarios and conduct pre-mission rehearsal.
The live orbital data integration is the most operationally significant architectural feature. If the training environment is seeded with real-world positional and behavioral data rather than scripted approximations, the scenarios a Guardian rehearses are closer to what they will actually encounter. That is meaningfully different from simulated training built on historical cases or notional threat representations.
The gap between that design intent and verified performance is worth tracking carefully. Whether TALOS can accurately model adversarial satellite behaviors—particularly deception-oriented maneuvering, proximity operations that approach but don't cross established thresholds, or electronic warfare effects—depends on the quality of its training data and the depth of its orbital dynamics models. Slingshot's capability claims are vendor claims. STARCOM testing evaluated TALOS for potential value, not verified operational readiness. These are different things, and the distinction matters before the platform becomes a primary readiness mechanism.
The capability questions planners should track:
- Threat model currency. How quickly does TALOS's adversarial model update as actual adversary behavior evolves? A training system that lags real-world threats by twelve months produces rehearsal that is calibrated against yesterday's threat.
- Consequence modeling fidelity. Rehearsal improves decision quality only if the system can model the consequences of different courses of action with sufficient accuracy. Generating plausible-seeming outcomes isn't the same as modeling realistic operational consequences.
- Independent assessment timeline. The program needs an evaluation protocol that tests model fidelity independently, not against Slingshot's own metrics. That protocol should be part of the OTTI program plan.
The Home-Station Architecture and What It Requires
The most consequential design choice in MENTAT is home-station access. Rather than scheduling rehearsal time at a dedicated simulation facility, Guardians would reach AI-enabled scenario training from their regular work environments.
The operational logic is sound. Readiness should be continuous. Pilots use simulators regularly. Cyber operators train in practice environments daily. Space operators have had limited analogous access. If Guardians can run protect-and-defend rehearsals from their duty desks between missions, readiness becomes a daily practice rather than a scheduled event. The MENTAT design is trying to solve a real structural problem in how space operations training works.
Home-station access implies requirements that need to be planned against, not assumed away:
Classification ceiling. The classification of meaningful orbital data—threat characterizations, specific tracking data, intelligence assessments of adversary behavior—constrains what MENTAT can deliver at home-station classification levels. A training environment that omits the most operationally important threat information because it can only reach NIPR or lower SIPR tiers won't produce the readiness improvement the program is designed for. The Space Force needs to define, early in the program, what classification level MENTAT will operate at, and whether that ceiling is compatible with the training fidelity the program requires.
Network requirements. Home-station access for space operations units—including forward-deployed or austere locations—requires adequate network capacity to reach a cloud-hosted AI training platform. That is a solvable problem for garrison-based units; it is a planning question for any unit that operates from degraded or limited connectivity environments.
Allocated training time. Making home-station access available doesn't automatically produce readiness gains. Squadrons need to actually schedule and use the platform. That is a workforce and organizational design question for STARCOM, not a technology problem, and it's the kind of implementation detail that determines whether a training investment produces measurable readiness improvement or sits underutilized.
What Planners Should Do With This
The MENTAT investment signals where Space Force training infrastructure is heading. Several implications are worth acting on rather than observing:
Use the OTTI footprint to shape requirements. MENTAT's architecture will influence how the Space Force specifies future training systems. If operational squadrons engage STARCOM and the OTTI program office with feedback about classification needs, scenario gaps, and interface requirements early in the program, that feedback has a chance of influencing the system. That window is open now; it will narrow as implementation matures.
Map the vendor dependency. The SBIR Phase III pathway produces working capability quickly, but it means the core of MENTAT is commercial intellectual property. Program planners should understand what modifications the government can direct without contract amendments, what data rights the Space Force has, and what continuity of operations looks like if Slingshot's circumstances change over a 4.5-year performance period.
Plan the independent evaluation protocol. The program needs a clear framework for evaluating TALOS's actual fidelity against operational requirements—not Slingshot's own performance claims—before MENTAT becomes a primary readiness pathway. Building that evaluation into the program plan now is less expensive than retrofitting it after the platform is fielded.
The direction of this investment is right. AI-enabled continuous mission rehearsal is a better model than episodic exercise access, and live orbital data integration is the correct foundation for it. What the Space Force needs from here is the operational discipline to verify fidelity, manage classification thoughtfully, and hold the vendor accountable to performance that actually translates to better-prepared Guardians.
Spartan X's work with AI system verification—specifically evaluating multi-model AI outputs under variable conditions before those systems are trusted in high-stakes operational contexts—is directly applicable to the challenge of certifying that an AI training platform like MENTAT produces calibrated, operationally useful rehearsal rather than plausible-but-misleading scenarios. Getting that verification right is what turns the investment into actual readiness.
Sources and Further Reading
- Slingshot Aerospace. "Slingshot Aerospace Wins $69.2 Million U.S. Space Force Contract to Advance AI-Powered Mission Readiness for Space Defense." BusinessWire, July 15, 2026.
- Slingshot Aerospace. "Slingshot Aerospace Awarded $27 Million Space Force Contract to Power the AI-Driven Training Environment for Space Warfare." BusinessWire, January 15, 2026.
- U.S. Space Force. Space Training and Readiness Command (STARCOM). spaceforce.mil/About/Organizations/STARCOM
- Defense Post. "Slingshot Aerospace Wins Space Force Contract for AI Training." The Defense Post, July 21, 2026.
