0>1.software

// digest/2026-08-20

August 20, 2026

Daily Digest: August 20, 2026

An AI-narrated daily AI and defense-tech briefing. Every item links to its source.


Army cyber chief reveals AI task force building agents to ‘hunt’ in the DOD network

Lt. Gen. Christopher Eubank described an Army task force building AI agents to hunt threats on DOD networks. He said the team meets daily to decide which risks a human should own versus what an agent can handle, and that humans currently retain all responsibility for risk; no agents have been given autonomous authority.

The daily human-in-the-loop ritual around risk ownership stands out more than the hunting agents themselves. It's the kind of guardrail most commercial agentic deployments skip until something breaks.

New White House strategy clarifies military tech priorities: undersea, outer space and AI

The White House's new National Security Science & Technology Strategy (NSSTS) names undersea, outer space, and AI as top military tech priorities. It also flags a second tier of longer-term technologies to watch, including quantum and biotech.

Naming three top priorities plus two longer-term watch items reads like an early signal for where DOD R&D dollars will flow next budget cycle.

Striveworks to build AI layer for NGC2 at corps level, Army says

The Army says Striveworks will build the AI layer for NGC2 at the corps level. CEO Jim Rebesco said the company's models can compress hours of staff work into seconds.

Compressing staff work into seconds is the same pitch every agentic vendor makes. Whether NGC2 integrates it deeply enough to change decision cycles, or just bolts on a chat interface, is still an open question.

Pentagon rebuffed repeated requests for Golden Dome brief: CBO

CBO National Security Directorate head David Mosher said the Pentagon repeatedly declined requests to brief the office on Golden Dome. He said CBO's task was to estimate the cost of implementing Trump's January 2025 executive order. He added that reviewing DoD's actual plans for the program was outside that task.

A federal budget watchdog getting stonewalled on a program's basic scope is a bad sign for cost transparency going forward.

China’s Military Says AI Can’t Replace Commanders. Xi Is Testing That

China's People's Liberation Army Daily has long insisted that AI can sort sensor data and draft options but that final command authority and responsibility stay with the human commander. War on the Rocks reports Xi is now testing whether that promise holds.

China's official doctrine sounds identical to the US line on human-in-the-loop control. Whether battlefield practice stays aligned with that doctrine under combat pressure is the harder test.

Nvidia Tries to Spur the AI Boom Into the Future

Bloomberg reports Nvidia's H200 chips are beginning to reach China. The AI race is broadening into a global contest over chips, electricity, factories, robots, and cybersecurity.

Nvidia's H200 chips reaching China shows the AI race broadening past raw chip specs into electricity, factories, robots, and cybersecurity.

War: Watt Is It Good For?

War on the Rocks examines how unmanned systems in Ukraine, aided by onboard AI, have captured infantry units and struck deep into areas where Russian jamming foils remote piloting, democratizing military capability for smaller countries. The piece notes this democratization runs alongside a consolidation visible once you strip down the underlying hardware.

That democratization comes paired with a consolidation underneath the hardware, the piece notes, once you strip the tools down to their components.

Z.ai's GLM-5.3 API goes live for all coding plan subscribers

Z.ai's GLM-5.3 model, built for coding, defensive cybersecurity, and long-horizon agentic tasks, is now generally available to all GLM Coding Plan subscribers. It ships at the same price as GLM-5.2.

Shipping GLM-5.3 to all coding plan subscribers at the same price as GLM-5.2 keeps Z.ai's pricing flat while adding coding, defensive cybersecurity, and long-horizon agentic capability. Real agentic benchmark results will determine whether the upgrade holds up.

// subscribe

Get the digest in your inbox

One email each weekday. The stories that matter, with a take. No spam, unsubscribe anytime.