April 2026 / USSOCOM TE 26-2
USSOCOM TE 26-2: validated at TRL 6, recommended for deployment.
Hypha was selected by PEO-SDA for USSOCOM Technical Experimentation 26-2 in April 2026. The published assessment report (May 1, 2026) recommended USSOCOM consider deploying AgentForge via OTA on CDAO Tradewinds within 60 to 80 days. This is what happened, what got validated, and what comes next.
AgentForge participated in USSOCOM Technical Experimentation 26-2 (TE 26-2) Agentic AI in April 2026, hosted at Avon Park Air Force Range and sponsored by PEO-SDA. The published assessment report, released May 1, 2026, recommended USSOCOM consider deploying AgentForge via OTA on CDAO Tradewinds within 60 to 80 days. The platform was assessed at TRL 6: system prototype demonstrated in a relevant operational environment.
What PEO-SDA validated.
The published PEO-SDA assessment validated five Hypha capabilities in writing:
- Swarm of autonomous AI agents covering contextual understanding, model building, semantic fusion, and visualization.
- On-the-fly custom ML model generation at the edge. PEO-SDA characterized this as "a significant departure from traditional ML workflows."
- Multi-modal data ingestion across full-motion video and drone footage, sensor signals, and operational reference material such as SOPs and ROEs.
- Designed for secure, disconnected edge deployments.
- Integration with existing SOF architecture.
The recommendation language for external use is taken verbatim from the report. The pathway named in the report is OTA via CDAO Tradewinds within 60 to 80 days. A technical interlocutor, J24, met with the AgentForge team during the assessment.
USSOCOM should consider deploying AgentForge because it significantly accelerates the intel cycle in time-sensitive mission threads. PEO-SDA assessment report, May 1, 2026
What is TE 26-2?
USSOCOM Technical Experimentation (TE) is the command's recurring exercise for assessing emerging capabilities against real Special Operations Forces requirements. TE 26-2 ran in April 2026 at Avon Park Air Force Range, sponsored by PEO-SDA, and focused on agentic AI for the SOF intel cycle. Participating vendors brought their platforms against a fixed corpus and a fixed operator workflow; PEO-SDA assessed each against operationally meaningful criteria.
Inclusion in TE is not vendor marketing. It is a structured assessment by the command's technical office against the actual problem space the command is trying to solve. The output is a written report that program offices, J-codes, and program executive offices read.
Why AgentForge was selected.
PEO-SDA selected AgentForge because the platform's three structural commitments match what the command is pulling for: operator-built (the staff officer drives the platform, not a vendor cycle), local-first (the platform runs in the room with the data, no cloud egress), and honest about its own failures (the platform refuses to deploy models that fail the holdout).
The selection and the subsequent assessment confirm that operator-built, local-first agentic AI is the shape USSOCOM is pulling for. The validation is documented in a published government report.
Demonstration areas.
Demonstration 1
Contextual understanding.
Hypha ingested doctrine, SOPs, ROEs, and historical after-action reports. Staff-officer natural-language queries returned cited, structured answers grounded in the corpus, with the source documents available for inspection.
Demonstration 2
On-the-fly model build.
Hypha ran autonomous EDA, feature engineering, model selection, training, evaluation, and deployment verdict against the operational telemetry corpus. The honest-evaluation surface refused to deploy a model that scored 0.953 in cross-validation and 0.143 on the holdout.
Demonstration 3
Multi-modal fusion.
Hypha integrated full-motion video, drone footage, sensor tracks, and document corpora into a single operator view with provenance per layer. The fused view answered queries the individual layers could not.
Operational requirements met.
The assessment evaluated Hypha against the operational criteria that matter for a deployed SOF capability: local-first execution on operator hardware, multi-modal data ingestion, autonomous on-the-fly model generation, integration with existing SOF architecture, and disconnected edge deployment. Hypha met every criterion in the published report.
Key capabilities demonstrated.
- Local-first execution. End-to-end pipeline on a 48 GB workstation, no cloud dependency.
- Autonomous EDA and model build. Operator natural-language query to dashboard, end to end.
- Honest evaluation. DO NOT DEPLOY verdicts where the holdout failed.
- Multi-modal fusion. FMV, drone footage, sensor tracks, document corpora in one view.
- SOF architecture integration. Compatible with existing operator workflows and reference materials.
Significance.
Selection for TE 26-2 was the entry point. The published PEO-SDA assessment was the outcome. AgentForge now holds a government-authored, named recommendation to deploy, from the program office that selected it. That is the strategic significance: TRL 6 is no longer self-declared; it is documented in a published USSOCOM assessment, with a vehicle (CDAO Tradewinds), a timeline (60 to 80 days), and a named technical interlocutor.
Open the conversation.
If your program office is sitting in the same problem space, the next step is a phone call. Bailey reads every inbound inquiry himself.