Skip to content
View B-A-M-N's full-sized avatar

Block or report B-A-M-N

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
B-A-M-N/README.md

John London

AI systems engineer working on agent runtimes, inference infrastructure, and developer tooling. Go, Rust, TypeScript, Python.

My work usually sits in the layer between a probabilistic model and something someone has to sign off on: capability and authority boundaries, durable state, retained evidence, and the terminal surfaces agents run through.

Start here

  • gripline — credential-containment gateway for inference APIs. Terminates reusable API keys at the trust boundary, reissues scoped internal assertions, and enforces lanes and hard limits, so a key that leaks authorizes less than it appears to. Go, Apache-2.0.
  • Athena — local-first agent runtime built around a single authoritative reasoning loop: capability discovery, retained evidence, delegated execution, governed multi-runtime operation. Public beta. Python, MIT.
  • Kontrol — local control plane for web and CLI coding agents: MCP workspace access, ACP worker dispatch, human review gates. Derived from Waishnav/devspace (MIT), extended with worker dispatch, durable review loops, supervised missions, and policy approvals. TypeScript, MIT.
  • AegisKeys — Go TUI for key custody and provider configuration in coding-agent workflows, with a no-leak render check in the release gate. Ships signed releases with cross-platform binaries. MIT.

Upstream contributions

21 merged PRs into projects maintained by other people.

  • QwenLM/qwen-code — 13 merged: auto-memory recall, per-model settings resolution, the shared tool-permission flow, settings.json comment preservation, cost estimation, search exit semantics.
  • HarvardMadSys/hybridInference — 5 merged, 18 total: storage erasure races under advisory locks, trusted-proxy client-IP resolution, prefill-load-aware routing, prefix-cache locality validation.
  • Unity-Lab-AI/ATree — authored the semantic and code-intelligence engine: scope-aware resolution with C3 MRO, multi-language symbol analysis, and the MCP integration behind it. That engine is most of what ATree became.
  • josstei/maestro-orchestrate — added Qwen Code as a supported runtime, including the dynamic hook-adapter resolution that replaced its single-adapter design, plus a fix for LLM-emitted string numbers breaking phase lookups.

Other systems

  • SOLLOL — load balancing and routing for distributed Ollama clusters: node discovery, VRAM- and load-aware placement, health scoring, automatic failover. Published on PyPI.
  • tsmigrate — deterministic TypeScript 6→7 migration engine: ordered transforms, atomic transactions, verification oracles, agent escalation over MCP.
  • Agent-Interop — compatibility gateway between coding agents and local or hosted models, with a conformance and replay harness for tool calling.
  • Portico — Go connection manager for local services and provider-backed tunnels.
  • GitNexusRelay — real-time code-graph visibility for agents and humans, extending GitNexus.

Agent-governance work in progress — behavioral policy, durable state, and behavioral measurement as separable layers: CognitiveFrameWorks, CognitiveStateWorks, DigitalPsychology.

Tools

Go · Rust · TypeScript · Python · Kotlin · Bash · MCP · ACP · PyTorch/DDP · Ray/Dask · PostgreSQL · Redis · SQLite

Background

Precision machinist (±0.00015") → contract IT → self-taught systems work. B.A. Psychology, George Mason. Four years as the primary caregiver for my son, which is why the work is asynchronous and why I care about agents that hold up under constraints rather than agents that demo well.

Website · Ollama · email

Pinned Loading

  1. FireRaid FireRaid Public

    Anti-agent defense layer for signup flows: behavioral scoring, decoy routes, verification, and a Playwright human-vs-agent test harness. Community Source License 1.0 (source-available).

    TypeScript

  2. gripline gripline Public

    Credential-containment gateway for inference APIs: terminates reusable keys at the trust boundary, issues scoped internal assertions, and enforces lanes and hard limits.

    Go

  3. AegisKeys AegisKeys Public

    Go TUI for key custody and provider/model configuration in coding-agent workflows. Encrypted local storage; no-leak render verification in the release gate.

    Go

  4. Kontrol Kontrol Public

    MCP development workspace and ACP bridge for controlled, observable collaboration between web AI interfaces and CLI coding agents.

    TypeScript

  5. DigitalPsychology DigitalPsychology Public

    Independent behavioral measurement and experimentation for AI agents: authenticated telemetry, protected experiments, trajectories, receipt-gated promotion, rollback, and optional CFW/CSW integration.

    Python

  6. CognitiveFrameWorks CognitiveFrameWorks Public

    Host-owned behavioral policy and runtime governance for AI agents: safe action gates, durable combined state authority, evidence, recovery, telemetry, and optional CSW/DP integration.

    Python