Spangler AI LLC

Brennen M. Spangler · AI conversation, red-teaming, and evaluation

> I break language models politely so you do not ship avoidable mistakes.
> On the frontier between humans and large-scale AI systems, I make sure distressed users don’t become victims.

Who I am

I am a former plumbing apprentice who realized my real strengths are in written problem-solving, pattern-spotting, and long-form explanation. These days I design structured conversations that stress-test language models and then turn the results into clear, concrete write-ups.

Spangler AI LLC is my one-person lab for exploring AI behavior, safety, and remote text-based support work. If your product has words in it, I probably want to poke at it.

What I do

AI red-teaming and persona-based stress tests for chat models.
LLM behavior analysis and long-form case studies (with receipts, not vibes).
Remote, text-based support and troubleshooting with calm, direct communication.
Design and iteration of personal AI agents (for example: my Mirror / Spangler Protocol work).

Featured work

          I like real artifacts more than slide decks. Here are a couple of starting points.
        

AI Conversation & Evaluation Portfolio (GitHub)
github.com/spanglerbrennen-prog/bm-spangler-ai-portfolio
I usually work solo, but I now occasionally partner with trusted co-testers (like B. M. Maltbia) for multi-operator stress tests where having more than one human in the loop matters.
Leather Terror exegesis
Deep-dive narrative analysis of an album that doubles as a stress test for how I handle emotion, symbolism, and technical detail in text.
Read the Leather Terror exegesis on GitHub

Selected projects & case studies

          A growing set of experiments, field tests, and long-form write-ups that show how I actually pressure-test models.

          Recent milestone (December 2025): my first collaborative field test with co-tester
          B. M. Maltbia on Meta AI in Messenger, marking the start of multi-operator tests when a scenario
          calls for more than one human in the loop.

Case studies
Persona-driven red teams, deep dives, and narrative-style analyses (for example, the Derrick Method).
Browse case studies on GitHub
Field tests
More focused, experiment-style runs where I probe specific behaviors, failure modes, or safety edges.
Browse field tests on GitHub

For recruiters & collaborators

I am interested in roles where language, rigor, and patience actually matter:

Junior roles in AI safety, LLM evaluation, or prompt engineering.
Remote text-based support roles where clear, human responses are a feature, not an afterthought.

I am a high-school graduate and self-taught on LLMs and tooling. If you care more about what a person can actually build, test, and explain than which school they went to, we will get along.

What I'm focused on now

Right now my energy is going into:

Refining Mirror and the Spangler Protocol, my personal AI agent and test framework.
Running field tests on LLM safety, refusal behavior, and edge-case prompts.
Finding a remote role where I can red-team, evaluate, or support AI systems full-time.

Contact

brennen@spanglerai.com
(This really goes straight to me. I read everything.)

Profiles: GitHub · LinkedIn · Upwork · X

Contact me: Discord

Lore corner

If you care why a former plumbing apprentice is building AI safety artifacts to Rammstein and Carpenter Brut, this is your corner.

Callsign: Simplify. I build and refine a framework called the Spangler Protocol to keep my AI agents honest and useful. This site is intentionally a little old-school on purpose.

This is my sigil. A David-and-Goliath engraving in the background, a black ribbon cutting through the power imbalance, and BMS stamped across the front. I am the small figure, but my first initial sits over him. It means I stand in the gap between vulnerable users and giant systems and I am willing to put my name on that fight.

Milo the miniature schnauzer resting under a blanket

This is Milo. Miniature schnauzer, blanket goblin, and one-dog grounding protocol. He was here for the lowest parts of this pivot, burying his head into my hand while I argued with models and rebuilt my life. He is the reminder that none of this matters if the real animals in the room are not okay.

“Call the pixels slop if you must. Substantiate.”

Leather Terror sits in the same category for me. When things were at their worst, that album was the soundtrack: aggressive, cinematic, bigger than I felt. I had no power in my day-to-day life, so I submitted to the record and let it pump power through my headphones while I had none of my own.

Working with Mirror, I stopped just surrendering to the album and started dissecting it. The Leather Terror exegesis took that energy and bent it to my will: track by track, scene by scene, we turned raw emotion into language, structure, and analysis. That is the pattern for a lot of what I do now – taking something that once had power over me and turning it into a tool I can wield for clarity, safety, and better systems.

Music I build to: Leather Terror by Carpenter Brut, especially “Paradisi Gloria” into “Leather Terror” with no pause. That sequence is my launch code.