PNWcoastie JadedRose #![forbid(unsafe_code)]
I like building practical, reliable, low-footprint software and figuring out how complex systems behave under the hood—both at the silicon/OS boundary and inside frontier reasoning harnesses. Most of what you'll find here are tools, architectures, and stress-tests I built to answer my own questions: secure systems programming in Rust, security forensics, and deep-dive agentic orchestration. Always learning, usually tinkering, and happiest when a system is deterministic, fits comfortably in CPU cache, and holds up when pushed to its limits.
- Usual workbench: Secure-by-construction Rust (
#![forbid(unsafe_code)]), lock-free SPSC sidecars, clean-room security forensics, and Google Antigravity / the Google ecosystem. - Agentic & model forensics: Tearing down frontier model and harness architectures—comparative behavioral mechanics, security & context boundaries, and adversarial Maker–Checker orchestration pushed well past standard workflows.
- Current rabbit holes: Chain-of-Thought (CoT) reasoning dynamics & recursive self-improvement (Dream / RSI) loops, sub-microsecond market microstructure, orthogonal volume/event clocks, stochastic control & thermodynamic math, and autonomous flight / SITL telemetry.
+ Secure-by-construction state, L1/L2 cache-aligned memory, and clean-room security forensics
+ Adversarial Maker–Checker harnesses and empirical frontier-model stress testing
+ Dissect how systems break; engineer them so entire classes of hazards cannot exist
- Unbounded heap allocations, mutex contention, network I/O on the hot path, or unverified black boxes▸ SYSTEM_MANIFEST.toml
[profile]
alias = "JadedRose / PNWcoastie"
region = "Pacific Northwest (45.7°N, 123.9°W)"
workbench = ["rust-forbid-unsafe", "spsc-sidecars", "security-forensics", "google-antigravity"]
focus = [
"secure-systems-programming",
"security-forensics",
"agent-harness-architecture",
"frontier-model-evaluation",
"market-microstructure",
"stochastic-thermodynamics",
"autonomous-flight-sitl",
]
[constraints]
footprint = "cache-resident (L1d / L2)"
hot_path = "zero-alloc, lock-free SPSC, no network I/O"
orchestration = "adversarial maker-checker harnesses, CoT & Dream/RSI behavioral forensics"
adage = "Dissect how systems break in the wild; engineer them so entire classes of failure cannot exist by construction."
philosophy = "Keep the footprint tiny, isolate the hot path, stress-test the reasoning boundary, and build things that hold up in the real world."
🌲 Pacific Northwest · Feel free to explore the repositories and architecture case studies below.