4‐3 Agentic AI industry examples - terrytaylorbonn/auxdrone GitHub Wiki
26.0411 (0317) Lab notes (Gdrive) Git
26.0411 https://youtu.be/tao-L5N0m6o?t=605
https://www.youtube.com/@webb8020 (Webb videos)
Best partner Youtube video "The Era of AI Warfare Begins": "What would the world be like if the outcome of war no longer depended on the number of tanks, but on the speed of algorithm iteration? This video takes you behind the scenes to uncover the truth behind Operation Fury, the epic operation that shocked the world in 2026. We will dissect the Silicon Valley kill chain comprised of Palantir, OpenAI, SpaceX, and Anduril, to see how AI can compress 100 days of work into 90 minutes and complete global cross-domain command synchronization within 200 milliseconds. When Silicon Valley's "software thinking" collides with the steel behemoth of the Pentagon, this ultimate game of technological sovereignty, ethical red lines, and geopolitics officially ushers in a chaotic era where algorithms define power.
GPT likes my analogy. Interesting conversation in Lab notes docx #600-2 (Gdrive).
The key: ontology. Since the UFA is an approximator, we lessen the chance of errors by limiting the world view. We make reality more granular. https://www.supplychaintoday.com/palantir-ontology-overview/
https://lnkd.in/gbVmQ-Ck. "Two operations in early 2026 shocked global military experts: the Caracas raid and Operation Epic Fury. This was not only a victory for AI, but also a complete rewriting of the rules of warfare. But you may not know that behind this AI war, the real winner wasn't OpenAI or Anthropic, the developers of the models, but the mysterious company that provided the underlying "skeleton"—Palantir." https://www.linkedin.com/posts/activity-7436522849521115137-WpNX
I would classify this as an eRobotic AI that
- Can make mistakes. This means that even if its not perfect, its still a viable product (EXCEPT for non-nuclear forces). The business case here is war. And in war there are mistakes, friendly fire, etc. The main point is: Would this system improve the survival of friendly forces and the destruction of hostile forces?
- Must be constantly under human observation (just like driving a Tesla). You cannot allow it to unsupervised like an airliner copilot.
It will directly be responsible for commands sent to weapons API (so its critical infrastucture such as databases, etc).
The following is just a wild guess at to how this is supposed to work.
- (1) Central command monitors intel.
- (2) Central command sends request to AI (secure foundational model) for battle plan (or maybe field command first sends suggested operation to central).
- (3) AI model/agent works with central/field command to to create plan.
- (4) Human (field or central command) presses the red button to execute strike.
- (5) C2 sends commands to weapons APIs.
What I wrote above is some wild guess. The only things that I sure about
- The Fury Agent is a CPU based serial processing script / code.
- The NN (UFA) is a foundation model for this battlefield scenario. THIS (along with the agent) IS THE KEY COMPONENT.
- A human will press a button (not a voice single) to initial battle (and to modify/stop). A human must always be in control, just like in a Tesla.
It has to be something like this because of the
- Inherent limitations of UFA-based "AI"
- The nature of battle
Really nice screenshot below. I would image this is not from a working system. In any case, the most important thing would be the performance of the UFA and the agent.
Just like AI agents are transforming the digital world, Fury Orchestrator brings that same agentic intelligence into the physical world for the U.S. warfighter. ME: AI agents only simulate intelligence. Based on UFAs. That is a CRITICAL distinction.
It is the first agentic system that can take a commander’s high-level mission intent, spoken or typed in natural language, and turn it into coordinated autonomous action across a mixed fleet of unmanned assets. ME: This will be very effective. It is not deterministic, but does not have to be. The main thing is it will get a very good solution (in much shorter time than a human could), and CAN MASSIVELY SCALE).
The following is a fully autonomous demo mission using AI agents at our proving grounds. We task Fury Orchestrator to use available unmanned assets (i.e. an unmanned ground vehicle and 2 one way attack drones) to find, fix, and destroy an enemy truck. ME: What the AI will return is what it has been "trained" on. So this will require a "foundational" model (a custom made model) for warfighting.
This whole mission was done handsfree using commander intent and AI agents, with human operators kept in the loop strictly for safety supervision. Important to note this was not pre-programmed conditional logic or decision trees, it was dynamic translation of mission intent into autonomous action. ME: If you are using an LLM, you are dependent upon the LLM understanding you. The human operator in any case will ALWAYS press some red button for instigate attack. You will NEVER do this with voice commands.
Here is how it works at a high level:
A commander gives an objective through our command & control (C2) interface. From there:
1/ Fury builds a mission plan and submits it for human approval
2/ Once approved, Fury translates intent into structured instructions native to each vehicle’s API
3/ Fury monitors telemetry, video, C2 data, and each agents system reasoning to maintain a live common operational picture
4/ And when conditions change, it dynamically re-tasks assets in real time. ME: But this will have to be closely monitored by humans.
Fury Orchestrator acts as an AI interoperability layer, bridging the gap between C2 systems and unmanned assets, without modifying their underlying control stacks. The result is one software layer that can command fleets of heterogeneous robotic systems
It keeps the human at the center, but enables one operator to direct coordinated mass across autonomous systems. ME: Fury should simply suggest battle plans, and modifications; A human is SURELY ALWAYS IN COMMAND... just like anyone in a Tesla.