Skip to content
View Shawn5cents's full-sized avatar

Highlights

  • Pro

Block or report Shawn5cents

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Shawn5cents/README.md

Shawn Nichols Sr. — agent systems, evaluations, verification, and local inference

/* public proof */

01AgentSkinauthorSemantic data reduction for AI agents — an MCP-first system for pruning noisy API responses before they reach the model.
02Grok Build Legion EditiondownstreamxAI Grok Build community fork adding heterogeneous model routing, role-specific agents, independent verification, bounded repair, and provider discovery.
03agent-zero-nvidia-nimauthorNVIDIA NIM chat + embedding provider for Agent Zero with live model discovery and self-hosted endpoint support.

/* selected research systems */

01InvariantprivateAutonomous engineering runtime built around one rule: a builder does not certify its own work. Sandboxed execution, independent verification, repair loops, replay, and release gates.
02micro-cpp-harnessprivateNative C++20 coding-agent harness for local open-weight inference, verifier-gated execution, trajectory capture, recovery, and reproducible benchmark qualification.
03ParticleprivateRecursive agent runtime with scoped workers, persistent specialists, context isolation, durable goals, recovery, and orchestration state outside the transcript.
04Agent Mode Benchmark LabprivateOriginal evaluation program for coding, computer use, interactive products, games, and long-horizon agent work with locked briefs, clean workspaces, provenance, and deterministic scoring.
05CollectiveprivateDistributed AI control plane with one Scheduler authority, bounded capabilities, exact routing, signed evidence, independent verification, and safe recovery.

/* product systems */

01Link2Noteprivate productTurns links, videos, PDFs, audio, playlists, and channels into structured notes, transcripts, archives, and files; includes auth, billing, gated workflows, and staged multi-model study/review modes.
02OpenSwitchprivate productOpenAI-compatible marketplace that deterministically routes model + provider by capability, benchmark quality, latency, health, funding state, terms, and cost.
03Chirpprivate alphaPush-to-talk interaction layer for existing AI chats and human channels while preserving the underlying conversation's context, tools, and history.

Private systems are available by demo or granted repository access where appropriate.

/* showcase */

Invariant and Collective verification architecture micro-cpp-harness and Agent Mode evaluation workflow
01 Invariant + Collective — authority, bounded execution, independent proof 02 micro-cpp + Agent Mode — local inference, traces, controlled qualification, reproducible evals

/* operating background */

01Walmart cold chainManaged time-critical compliance operations where traceability, process discipline, service reliability, and failure recovery were hard requirements.
02Mercedes JITManaged just-in-time automotive logistics where sequencing, dependencies, timing, escalation, and recovery directly affected downstream operations.
03mechanism transferI learned systems in operations before I learned them in code: route the right resource, bound authority, preserve evidence, verify independently, and recover safely.

/* principles */

information != authority
claimed completion != verified completion
model choice != capability proof
failure data is part of the result

/* contributions */

Shawn5cents contribution calendar in 3D

Shawn Nichols Sr. · Independent Research Engineer — Agents, Evals & Local Inference · @Shawn5cents · agentskin.dev

Popular repositories Loading

  1. grok-build-legion-edition grok-build-legion-edition Public

    Forked from xai-org/grok-build

    Provider-flexible, role-based multi-agent DAG for Grok Build with model discovery, tool-call repair, verification, and fallback routing.

    Rust 5 1

  2. agentskin agentskin Public

    AgentSkin Suite: Complete token optimization stack — SSS JSON pruning + Tokenjuice CLI compaction + Caveman output compression. 88%+ API savings, 99.97% CLI savings, 4,695 tests.

    TypeScript 1

  3. Shawn5cents Shawn5cents Public

    Config files for my GitHub profile.

  4. agent-zero-nvidia-nim agent-zero-nvidia-nim Public

    NVIDIA NIM chat and embedding model provider plugin for Agent Zero

    HTML

  5. a0-plugins a0-plugins Public

    Forked from agent0ai/a0-plugins

    Agent Zero plugins index

    Python

  6. agent-engineering-skills agent-engineering-skills Public

    Production-grade engineering skills for Agent Zero — adapted from addyosmani/agent-skills

    Python