Software I built because the work kept repeating.

HG PPC

internal

ad data warehouse

165+ ad accounts modeled

x-accel

live

content scoring

public growth system

Numbers hub

live

public claims and caveats

claims linked to evidence

Open Agents

internal

agent fleet

80+ scheduled jobs

event log

rolling

/schema synced

/ad account scored

/field note indexed

/proof card refreshed

/agent handoff checked

Why this belongs here

This is the pattern: build the software, score the data, expose the public proof, and keep the notes where I can find them again.

The last six months

What I shipped while everyone else was tweeting about AI.

Each of these is real working software, not a slide deck. Some run publicly. Some run internally. None of them are demos.

Core

Shared software

HeavisideOS is the codebase, data layer, and agent fleet under the agencies. One fix can help more than one brand.

Landing

Niche packs

MarketingOS is what the team uses today. PavingOS and GarageDoorOS are next: same engine, screens shaped for paving and garage door owners.

Internal

Ad intelligence

Meta ad scanner across the niches, PPC data warehouse, creative review loop. Built so the kill/scale/refresh decisions stop being made on Friday vibes.

Live

Growth tools

x-accel for X. HG Market Report for new-client audits. HG Content for publishing. Cold email tools for outbound. I build what the agencies are missing.

Loading GitHub activity...

Shared software

HeavisideOS

Internal

The software my agencies run on. One codebase under the websites, CRMs, dashboards, ad pipelines, content tools, and AI agents. Different brands on top.

Result: One reusable core now sits under the agencies, dashboards, content systems, ad-data workflows, and agent fleet.

Built because: The old version of the business depended on scattered SaaS tools, docs, manual reporting, and Monday memory.

30+ production systems running on it

Powers MarketingOS, PavingOS, GarageDoorOS

systems:heavisideos

Niche software

MarketingOS

Internal

The cockpit my agencies use day to day. SEO, PPC, content, websites, CRM, reporting, outbound, wired together so a Monday review takes 30 minutes instead of three hours of tab-switching.

Result: Agency growth work is organized as one loop: outbound, ads, SEO, content, websites, CRM, reporting, and owner review.

Built because: Running growth from separate ad dashboards, GHL screens, spreadsheets, and notes made every review slower and less trustworthy than it needed to be.

Runs Heaviside Digital, Paving Marketers, Garage Door Marketers, EMA

Feeds patterns into PavingOS and GarageDoorOS

systems:marketingos

Niche software

PavingOS

In Progress

A version of MarketingOS shaped specifically for paving contractors. Built from what I've learned running Paving Marketers since 2020.

Result: PVM is where the paving-specific software has to prove itself, not just an agency case study.

Built because: Paving owners need estimate follow-up, commercial pipeline visibility, job economics, and local search clarity more than generic marketing reports.

Niche refresh in progress

Built on live Paving Marketers operations

systems:pavingos

Niche software

GarageDoorOS

In Progress

MarketingOS shaped for garage door installers and repair shops. Garage door is fast. Emergency calls, same-day jobs, brutal competition. The software has to keep up.

Result: A second niche proves the HeavisideOS pattern can work outside paving.

Built because: Garage door companies deal with urgent demand, missed calls, emergency scheduling, stale estimates, and franchise pressure.

Niche refresh in progress

Built on live Garage Door Marketers operations

systems:garagedooros

Revenue tools

HG Websites

Live

The multi-tenant Next.js platform under the agency sites and most client sites. One codebase, many brands.

Result: One website engine now powers the public web layer for the Heaviside portfolio instead of rebuilding the same pages, forms, tracking, and proof blocks for every niche.

Built because: Every new agency/site needed the same pages, tracking, forms, proof blocks, and Core Web Vitals work.

Powers Heaviside Digital, PVM, GDM, EMA, Heaviside AI, Heaviside Group

Animation and interaction refreshes in progress

systems:hg-websites

Revenue tools

HG Web Commander

Pilot

The HeavisideOS-driven runtime that builds client-niche sites from templates, runs preview and approval flows, and records rollback evidence.

Result: Client-site production gets a governed preview, QA, publish, and rollback layer instead of a manual handoff process.

Built because: Website refreshes fail at the handoffs: review state, domain readiness, publishing, rollback, and evidence.

Pilot deployment for niche refreshes

Bridges HeavisideOS into client-site rendering

systems:hg-web-commander

Revenue tools

Agency Financials

Live

The financial dashboard for the agencies. Revenue, costs, margin, client economics, forecast. One screen instead of QuickBooks plus three spreadsheets.

Runs the Heaviside agencies on a 35/20/20/25 cost structure

Q1 2026 cashflow model: 24 days actual, zero drift

systems:agency-financials

Revenue tools

Cold Email Engine

Internal

The outbound stack: list building, verification, mailbox infrastructure, send pacing, reply triage, suppression, reporting. Built like a system so it stops being a quarterly scramble.

Supports Heaviside agency growth

Feeds Digital Agency Growth Academy curriculum

systems:cold-email-engine

Revenue tools

HG SEO Commander

Live

The screen the SEO team works from. Audits, visibility tracking, GBP posting, content queues, review prompts, and the AI visibility checks that started mattering when ChatGPT and Claude started recommending businesses by name.

Result: SEO work becomes staged actions, reviewable recommendations, and agent-assisted execution instead of a monthly PDF.

Built because: SEO delivery gets vague when work, evidence, and review state are disconnected.

30+ cataloged SEO/GBP/GEO agents (dry-run / staged execution)

Live across Heaviside agency clients

systems:hg-seo-commander

Revenue tools

HG Directory

Live

The Go + PostGIS engine that powers pavinglist.com, garagedoorlist.com, and electricianslist.com, owned-media directories with ~148K combined listings.

~148K listings across paving, garage door, electrical

Hermes-run moderation every 5 minutes

systems:hg-directory

Ad intelligence

HG PPC

Internal

A warehouse for Google and Meta ad data. Vertical labels, creative pattern tracking, ML feature views. An account-level scoring and alerts engine running in production.

Result: Paid-media review moves from dashboard-watching to benchmarked account scoring, risk detection, and waste/fatigue alerts.

Built because: Google and Meta dashboards show local activity but miss cross-account context, broken tracking, and silent deterioration.

165+ ad accounts across paving, garage door, and Heaviside verticals

Production ML scoring on every warehouse run

systems:hg-ppc

Ad intelligence

Meta Ads Intelligence

Internal

The Meta side of HG PPC, plus the creative review loop. Account health, fatigue scoring, competitor moves, and the kill/scale calls that decide the month.

Result: Meta creative decisions become a repeatable review process instead of vibe-based calls.

Built because: Fatigue, scale, and replacement decisions are expensive when they rely on memory and screenshots.

Live across Heaviside agency Meta accounts

Connected to competitor scanner and HG PPC warehouse

systems:meta-ads-intelligence

Ad intelligence

Facebook Ads Competitor Scanner

Lab

Scans Meta's ad library across my niches and turns it into something I can actually read on a Monday. 1,113 ads, 209 pages, 24 deep teardowns to date.

Result: Competitor ad research gets organized into market patterns that can feed actual creative strategy.

Built because: Manual Meta Ad Library research is slow, noisy, and hard to reuse.

1,113 ads collected

209 advertiser pages tracked

systems:facebook-ads-competitor-scanner

Ad intelligence

HG Market Report (software)

Live

The Next.js + DataForSEO engine that produces every HG Market Report. Local rankings, ad libraries, review signals, geographic coverage, competitor offers, packaged into a contractor-readable report.

Result: A market report that turns local rankings, ads, reviews, geography, and competitor offers into something an owner can actually read.

Built because: Most agency audits are generic PDFs that do not show the real market or the next decision.

Live product at hgmarketreport.com

Used as the audit step for new agency clients

systems:hg-market-report-software

Content tools

HG Content

Live

The pipeline that produces and ships content across the agency brands. Multi-brand planning, AI-assisted drafting, SEO workflows, and the editor pass that keeps it from sounding like everyone else's AI content.

~900 articles/month across client and internal sites

Part of the MarketingOS content layer

systems:hg-content

Content tools

x-accel

Live

The tool I use to plan, draft, score, schedule, and learn from posts on X. Same engine grew an anonymized account 22.7x in 90 days. Public SaaS at hgxaccel.com.

Result: A content growth loop that drafts, scores, schedules, publishes, watches results, and feeds the next round.

Built because: Audience growth is hard because the work is repetitive and the learning loop is easy to lose.

22.7x follower growth on an anonymized account in 90 days

383K impressions in the same window

systems:x-accel

Content tools

AI Video Pipeline

Live

Next.js + Remotion app for generating Meta and X video ads. ElevenLabs for voice, Whisper for transcripts, FLUX 1.1 / Gemini 3 Pro for stills, ComfyUI on a local RTX 3090 for the heavy lifting. ~$0.34 per finished 30-second ad.

Result: Video creative becomes a repeatable production pipeline instead of a slow vendor bottleneck.

Built because: PVM/GDM need more ad creative than a manual video workflow can produce economically.

~30 minutes per video, ~$0.34 cost per finished ad

Used for PVM and GDM Meta campaigns

systems:ai-video-pipeline

Agent infrastructure

Hermes Agent Fleet

Internal

56 scheduled jobs across three machines (HQ + two workers). Monitors infrastructure, queues, backups, anomalies, M365 inbox, follow-ups, day briefs, research content. The agent layer underneath everything.

Result: Agents run as monitored operations with schedules, budgets, approvals, and human checkpoints.

Built because: Agent demos do not handle wakeups, retries, governance, logs, or approval routing.

56 scheduled jobs (50 HQ + 3 apps-vps + 3 homelinux)

MiniMax-M2.7 default, Z.AI GLM-5.1 for docs, gpt-4.1-mini for moderation

systems:open-agents

Agent infrastructure

Agent Commander

Open Source

Multi-host dashboard for AI agent sessions. Real-time terminal streaming, attention queue, scoped memory, approval flows. The open-source version of how I run agents.

Live at agentcommander.co

Open source: github.com/cvsloane/agent-commander

systems:agent-commander

Agent infrastructure

Infrastructure Dashboard

Open Source

Self-hosted dashboard for Coolify deployments, BullMQ queues, Prometheus, PostgreSQL health, the Hermes fleet, and the home-network gear. SSE updates so it's actually live.

Result: One dashboard shows the infrastructure the businesses depend on: deployments, queues, databases, agents, traces, and network health.

Built because: Production operations were spread across Coolify, shell commands, logs, queue views, Prometheus, and memory.

Live monitoring across apps-vps, db-vps, homelinux

Open source: github.com/cvsloane/infra-dashboard

systems:infra-dashboard

Open source

Open source tools

Open Source

The smaller public projects: MCP servers, OpenClaw skills, the Bitwarden Secrets initializer, the Mission Control fork. Tools I use that other people can use too.

10+ active public repos

Used inside Heaviside and shared outward

systems:open-source-tools