Digital Maker

The future is bright. Let’s build it.

We live in the golden age of the software maker. Armed with clear architecture, sharp primitives, and autonomous AI agents, one determined builder can ship today what used to take a whole department. I spend my days turning ideas into working code.

Creator of stat_pg in-engine PostgreSQL telemetry (1B+ MPS)
Architect of Simple AI autonomous delivery workflow
Author of MechaLinters deterministic guardrails for AI coding agents
Creator of MechaRuby security tooling
Creator of avp-mvd-menu-bar Apple Vision Pro menu bar utility
Maker of 33+ tiny open-source libraries
Independent game studio at games.jefflunt.com
Former top 1% contributor on Stack Overflow

Active Radar (Recent) Showcase Carousel (7) Workshop (33) statpg simpleai mechalinters MechaRuby avp-mvd Games Studio ↗ GitHub LinkedIn X / Twitter

BUILDER_MODE: ONLINE FUTURE // BRIGHT

“I build because it is the single greatest creative outlet I have ever discovered.”

Active Radar

Fresh off the Bench

The latest releases, tools, and platforms shipped across the workshop. Auto-synced on deploy.

Just Shipped Nov 2026

avp-mvd-menu-bar

Native macOS menu bar utility for Apple Vision Pro Mac Virtual Display pre-flight readiness and instant diagnostics.

1.147B MPS Oct 2026

stat_pg

In-engine PostgreSQL telemetry engine parsing 1.147 billion metrics/sec with eBPF XDP bypass and lock-free rings.

Zero Drift Oct 2026

MechaLinters

Deterministic AST auditors and compiler-grade CI attack probers eliminating stochastic drift for AI coding agents.

Autonomous Oct 2026

Simple AI

Minimalist high-leverage agent delivery pipeline chaining 5 atomic skills into the /shipit meta-skill.

Software is pure creative leverage. You start with an empty buffer and a problem. By afternoon, a new system routes packets, proves security invariants, or coordinates AI agents across a codebase. Nothing matches the thrill of shipping something real.

Cynicism is a trap for programmers. It is easy to complain about bloated stacks, sprawling microservices, and fragile dependencies. The cure is not complaint. The cure is building the clean alternative yourself.

AI agents multiply craftspeople; they do not replace taste. When you pair frontier models with small, readable modules and strict mechanical guardrails, the friction between imagination and working software vanishes. You stop fighting boilerplate and spend your energy on architecture, speed, and polish.

Keep the surface area small. Make the boundaries explicit. Put agents to work on the heavy lifting, verify everything in CI, and ship with a smile.

Every project on this page comes from that impulse: see a gap, open an editor, and build a sharp tool that solves it cleanly.

Three Rules from the Workbench

How I approach software architecture, AI tooling, and open-source libraries.

01 / SHIP WORKING CODE

Builders Over Committees

A working prototype teaches more in ten minutes than a specification document teaches in ten weeks. Write the code, run it against real inputs, and let reality guide the design.

02 / WELCOME THE AGENTS

High-Velocity Guardrails

AI lets small teams move at blistering speed. Pair that speed with closed-by-default registries, static AST auditors, and live probers so velocity never breaks safety.

03 / ZERO-BLOAT PRIMITIVES

Small, Sharp & Readable

Great libraries do one job and fit in your head. If you can read the entire source file in five minutes, you own the dependency instead of the dependency owning you.

Flagship Showcase

Major Systems & Platforms

stat_pg

statpg.com →

A billion metrics a second. Right inside PostgreSQL.

Stop running separate time-series databases, external aggregator daemons, and fragile synchronization pipelines just to count events. stat_pg embeds a StatsD-inspired high-speed metric ingestion engine directly into PostgreSQL as a native background worker.

Reaching 1.147 billion metrics per second requires relentless mechanical sympathy: eBPF XDP kernel driver bypass, lock-free double-buffered shared memory rings, and hierarchical cascading rollups that turn high-velocity telemetry into instant SQL queries.

eBPF XDP Driver Bypass: Intercepts UDP frames before kernel socket allocation, parsing up to 9.5M packets/sec and 13 GB/sec with 0% packet loss.
Lock-Free Double Buffering: Receiver workers ingest metrics into memory rings; atomic sub-microsecond buffer swaps ensure network I/O never waits on disk.
Native SQL Joins: Aggregations roll up cleanly into standard tables (stat_pg_secs, mins, hours, days), letting you join live telemetry with production tables in standard SQL.

postgresql.conf & psql SQL

# 1. Enable extension worker in postgresql.conf
shared_preload_libraries = 'stat_pg'
stat_pg.workers = 4
stat_pg.port = 8125

# 2. Query telemetry directly alongside production data
SELECT m.name, s.at, s.value
FROM stat_pg_secs s
JOIN stat_pg_metrics m ON m.id = s.metric_id
ORDER BY s.at DESC;

$ stat_pg benchmark --sustained BENCHMARK

Running eBPF XDP line-rate UDP stress test...
✓ 1.147 billion metrics/sec persisted to PostgreSQL tables.
✓ 13.03 GB/s wire throughput (9.50M packets/sec, 0% drop).
✓ 2.15 trillion metrics aggregated into 1.16 MB disk space.

Simple AI

simpleai.jefflunt.com →

A minimalist, high-leverage AI workflow that actually ships.

No complicated tooling. No fragile framework spaghetti. Simple AI pairs a frontier model and agent harness with five atomic skills—goal, clarify, rebuild, wrapup, and redeploy—chained into a single release meta-skill: /shipit.

Instead of drowning in orchestration boilerplate or babysitting brittle agents, the workflow enforces mechanical alignment, automated build and CI verification, semantic version tagging, and production deployment in a single continuous pipeline.

5 Atomic Skills: Chaining goal (persistence), clarify (intent check), rebuild (build/CI validation), wrapup (docs, version bump, commit/tag/push), and redeploy (production rollout).
Zero-Tolerance Gating: Executes release lifecycles end-to-end with zero tolerance for broken builds or failing tests.
Open-Source Tooling: Complete skill manifests and one-shot install prompt available at github.com/jefflunt/simpleai-skills.

pipeline.sh · /shipit orchestration WORKFLOW

# The 5-stage autonomous release pipeline
/clarify   # Verify intent & establish boundary assertions
/rebuild   # Compile, lint, and run project CI test suites
/wrapup    # Backfill docs, bump version, commit & git tag
/redeploy  # Push to production & execute health probes

✓ Executed inside an autonomous /goal loop

$ /shipit "deploy feature release" TERMINAL

Running Simple AI delivery pipeline...
✓ Clarify: Fast-path confirmed, zero blockers.
✓ Rebuild: 0 errors, all test suites passing clean.
✓ Wrapup: Version bumped, committed, tagged, pushed.
✓ Redeploy: AWS CloudFront & S3 synced. Live HTTP 200.

MechaLinters

mechalinters.com →

Deterministic guardrails & compiler-grade armor for AI coding agents.

AI coding agents can generate whole features in thirty seconds—and open five subtle architectural or security holes along the way. Prompting an LLM or relying on a secondary AI review bot to enforce macro-rules fails: you cannot cure stochastic drift with more stochasticity.

MechaLinters replace vague prompts with mechanical, binary truth: encoding complex architectural invariants, multi-tenant boundaries, and authorization matrices into sub-second AST auditors and behavioral CI attack probers.

The 4 Pillars: Single source of truth (declarative registries), closed by default, compiler-grade deterministic feedback, and behavioral CI attack probers.
Zero Stochastic Drift: Eliminates hallucinations and LLM review drift by giving agents line-numbered, actionable remediation contracts.
Reference Architecture: Battle-tested across production systems and powering the MechaRuby security ecosystem.

$ mechalinter verify DETERMINISTIC

FAIL: Missing tenant scoping in controllers/invoices_controller.rb:14
  → Route: GET /invoices/:id
  → Violation: Bare ActiveRecord query `Invoice.find(params[:id])`
  → Requirement: Must use `current_tenant.invoices.find(...)`
  → Status: Build blocked. Remediation contract generated.

ci.sh · Behavioral Prober SECURITY PROBE

Running MechaLinter architectural verification...
✓ 100% routes registered in authorization matrix.
✓ 0 un-scoped database queries across 48 controllers.
✓ AST audit passed in 18ms. Agent verified to green.

MechaRuby

mecharuby.com →

Rails + AI gives you speed-to-market. MechaRuby is the armor.

AI agents can ship features on day one, but open-by-default frameworks get breached on day two. One forgotten before_action exposes an endpoint. One bare Invoice.find(params[:id]) leaks another tenant's data.

MechaRuby puts Ruby on Rails in strict mode. Every gem combines a single-file YAML registry, a millisecond static AST or schema auditor, and a live behavioral prober in CI.

mecha_authorize: Maps every route and HTTP verb in config/mecha_authorize.yml. Unmapped routes return 404. Period.
mecha_tenant: Parses controllers with Ruby's Prism AST parser and fails the build if a query skips tenant_scope.
mecha_encrypt: Accounts for every database column, verifies active ciphertexts, and blocks broken indexes on non-deterministic columns.

config/mecha_authorize.yml YAML

routes:
  - route_pattern: "/messages/:id"
    verbs: ["GET", "PATCH", "DELETE"]
    public: false
    admin: false
    user_ui: true
    user_api: true

$ bin/rails mecha:verify CLI

Running MechaAuthorize static & behavioral persona matrix...
✓ 84 routes mapped, 336 role assertions passed, 0 bypasses.

Running MechaTenant Prism AST controller audit...
✓ 24 models registered, 76 actions scoped via tenant_scope.

Running MechaEncrypt schema & ciphertext verification...
✓ 198 columns audited, 51 ciphertext envelopes verified.

avp-mvd-menu-bar

github.com →

Native macOS telemetry and instant diagnostics for Apple Vision Pro Mac Virtual Display.

Mac Virtual Display turns the Apple Vision Pro into a massive, ultra-crisp private workstation. But when the wireless handshake stumbles, macOS leaves you guessing in System Settings while you toggle radios and hope for the best.

avp-mvd-menu-bar lives in your menu bar and continuously audits the entire pre-flight connection stack. Built in Swift and SwiftUI, it provides immediate visibility into every prerequisite required for a flawless spatial display link.

Pre-Flight Service Auditing: Continuously monitors Wi-Fi, Bluetooth, AirPlay discovery, and the macOS Keychain security subsystem.
Proximity & Peer Detection: Actively listens for nearby Apple Vision Pro broadcast beacons and confirms device identity before initiation.
One-Click Display Handoff: Triggers Screen Mirroring and spatial display handshakes directly from the menu dropdown with zero latency.
Adaptive Diagnostic Polling: Idles at 30-second heartbeats during nominal operation, automatically ramping to 10-second checks when diagnosing connection drops.

AVPMonitorService.swift SWIFT

// Real-time spatial display prerequisite observer
let preflight = MVDPreflightCheck(
  wifi: true,        // AWDL / Wi-Fi direct interface
  bluetooth: true,   // CoreBluetooth LE peripheral
  keychain: true,    // iCloud security identity
  peerNearby: true   // Apple Vision Pro detected
)

guard preflight.isReady else {
  statusBarItem.status = .warning(preflight.diagnosticHint)
  return
}
statusBarItem.status = .readyToConnect

$ avp-mvd status --verbose DIAGNOSTICS

Auditing Mac Virtual Display prerequisites...
✓ Wi-Fi: Interface en0 active (5GHz 802.11ax, AWDL nominal).
✓ Bluetooth: Controller online, BLE advertisement active.
✓ Keychain: Identity certificates verified (same Apple ID).
✓ Peer: Apple Vision Pro detected (RSSI: -42 dBm).
✓ Status: Ready for Mac Virtual Display connection.

Systems & Network Engines

Protocols · Shells · Concurrency

Understand the machine all the way down. Build the shell, the wire protocol, and the scheduler.

The best way to master distributed systems and concurrency is to build the primitives from scratch. Over the years I have built custom command shells, network message specification languages, distributed object brokers, and parallel I/O engines.

Each tool strips away accidental complexity to expose the clean mechanics underneath.

nobject & nmspec: Network-hosted objects that respond like local Ruby instances, backed by a declarative wire message specification language.
ractor_io & thread_io: True multi-core Ractor parallelism and multi-threaded I/O primitives for high-throughput Ruby workloads.
mmsh & tiny_tcp_service: A specialized shell environment and string-based TCP server foundation powering tiny_work_service and tiny_info_service.

distributed_worker.rb RUBY

require "nobject"
require "tiny_lb"
require "tiny_rl"

# Rate-limited, load-balanced remote object calls over TCP
limiter  = TinyRL.new(max_requests: 500, per_seconds: 1)
balancer = TinyLB.new(["10.0.1.12:9000", "10.0.1.13:9000"])

limiter.throttle! do
  node = NObject::Client.connect(balancer.next)
  puts node.compute_monte_carlo(iterations: 100_000)
end

Jeff Lunt Games

games.jefflunt.com →

Press start. No downloads. No bloat. Just play.

Modern games too often make players sit through endless cutscenes, battle passes, and fifty-gigabyte downloads before showing the first frame of gameplay. I build games that run directly in your browser with zero friction.

Jeff Lunt Games is an independent arcade and studio featuring eleven original and tribute titles—spanning real-time castle sieges, procedurally generated infinite labyrinths, 8-bit run-and-gun platformers, and deep-space mining simulations.

Instant Browser Play: Pure canvas, WebGL, and lightweight engines running at 60 FPS on desktop and mobile browsers.
Zero Installation: Click a link, read three rules, and start playing in three seconds.
Eclectic Studio Slate: Featuring King’s Claim, InfiniMaze, Recontra, Apex Asphalt, Keys, DVMO, and Asteroids Tribute.

arcade_manifest.yml ARCADE

studio: "Jeff Lunt Games"
url: "https://games.jefflunt.com/"
catalog:
  - title: "King's Claim"
    genre: "Real-Time Castle Siege & Strategy"
  - title: "InfiniMaze"
    genre: "Co-Op Infinite Procedural Labyrinth"
  - title: "DVMO"
    genre: "Deep-Space Asteroid Mining Simulator"
  - title: "Recontra"
    genre: "8-Bit Retro Run-and-Gun Action"

$ open https://games.jefflunt.com/ PLAY

Connecting to arcade cluster...
✓ 11 games loaded. WebGL canvas initialized (60 FPS).
✓ Zero downloads required. Press start to play.

Open-Source Catalog

The Maker’s Workshop

Thirty-plus focused libraries, shells, network services, and tools built for clarity and speed.

Sort:

mecharuby AI & Security

Closed-by-default route authorization, tenant isolation, and schema encryption for Rails.

avp-mvd-menu-bar macOS · Swift

Native macOS menu bar utility for Apple Vision Pro Mac Virtual Display readiness and troubleshooting.

tiny_gemini AI & Security

Zero-bloat Ruby gem for interfacing directly with Google's Gemini LLM.

tiny_pair AI & Security

Terminal-native pair programming harness powered by Google's Gemini.

tiny_chat_gpt AI & Security

Minimalist Ruby interface for OpenAI's ChatGPT API.

mmsh Systems

Specialized shell environment built for fast interactive command workflows.

nmspec Network

Network message specification language for deterministic wire protocols.

nobject Network

Network-hosted objects with transparent, local-like method calls.

tiny_tcp_service Network

Fast, string-based TCP server wrapper for building custom micro-daemons.

tiny_work_service Systems

Network job queue built on top of tiny_tcp_service.

tiny_info_service Systems

Lightweight system telemetry service powered by tiny_tcp_service.

tiny_lb Network

In-process load balancer for distributing requests across upstream targets.

tiny_rl Systems

Fast, zero-dependency rate limiter for Ruby applications.

ractor_io Concurrency

Ractor-based parallel I/O primitives for multi-core Ruby execution.

thread_io Concurrency

Clean multi-threaded I/O coordination for concurrent Ruby workloads.

tiny_bus Data

Minimalist publish/subscribe message bus for decoupled in-process events.

tiny_pipe Data

Composable, reusable data pipelines for multi-stage transformations.

tiny_ring Data

Bounded, fixed-memory ring buffer implementation in pure Ruby.

tiny_monte Simulation

Monte Carlo simulation engine for probabilistic modeling.

tiny_outcome Stats

Predictive statistical pattern matching and outcome estimation.

tiny_mem Profiling

Measure exact memory consumption deltas across Ruby code blocks.

tiny_dot Data

Clean JSON-like dot-notation access for nested Ruby hashes.

tiny_bar CLI

Text-based terminal progress bar with labels and ANSI color support.

tiny_color CLI

Colorize command-line output cleanly with zero-bloat ANSI escape codes.

tiny_eta CLI

Calculate accurate remaining time estimates for long-running batch tasks.

tiny_log CLI

Minimalist, high-clarity structured logging tool for Ruby scripts.

pretty_bigdecimals Formatting

Human-friendly BigDecimal formatting for Ruby financial and math output.

pretty_floats Formatting

Clean floating-point number formatting for Ruby.

pretty_integers Formatting

Readable integer formatting with clean delimiters for Ruby.

timecalc Web Tool

Instant browser-based timesheet and duration calculator.

remark Web Tool

Browser tool for converting pasted formatted rich text into clean Markdown in real time.

vim4evar Craft

Practical Vim documentation, motion reference, and tribute site.

xnk-player Web Audio

Zero-bloat Web Audio player component with real-time EQ visualizer and ID3 parsing.

No tools matched that filter. Press Esc to reset.

Pick a theme, change the vibe.

Press T anywhere on the page to cycle palettes, or toggle light and dark mode in the header.

Track Lyrics

The future is bright. Let’s build it.

Fresh off the Bench

avp-mvd-menu-bar

stat_pg

MechaLinters

Simple AI

The Joy of Making Things

Three Rules from the Workbench

Builders Over Committees

High-Velocity Guardrails

Small, Sharp & Readable

Major Systems & Platforms

stat_pg

Simple AI

MechaLinters

MechaRuby

avp-mvd-menu-bar

Systems & Network Engines

Jeff Lunt Games

The Maker’s Workshop

Pick a theme, change the vibe.