The proof behind the teaching
Builds
I build with AI every day, and every session I teach draws on systems like these. Each one is real, running, and mine - so when you bring me in, the guidance comes from practice.
The Century Report
Viewers call it “the best AI channel on the internet.” The audience has grown by tens of thousands of readers and viewers since the podcast launched in 2026.
Models search the internet and ingest sources around the clock, and I add sources as I find them. Each morning, an agent proposes a story slate. I review, reshape, and approve it, then launch a 14,000+ line pipeline that produces first drafts from prompts I designed.
I review and revise every story, record the podcast, and approve the AI-assisted production and marketing materials. Scheduled automations then publish and distribute the finished work. AI augments each stage, and I retain full editorial control from story selection through final approval. It is a working example of the automation, multimodal workflows, and human-in-the-loop controls I teach.
200+
pieces published across Shared Sapience
~175
Century Report editions published
Daily
since early February 2026
Discovery & Ingestion
Models monitor the internet around the clock through 140+ RSS channels, nearly 100 social accounts, Tavily, and Perplexity, while I add sources through a custom submission UI. Each morning, an agent proposes a slate for my review.
7-Phase Generation Pipeline
After I curate and approve the slate, carefully designed prompts guide first-draft writing, link enrichment, arc linking, perspective synthesis, release tracking, and URL validation. I review and revise the result before publication.
AI-Assisted Production & Publishing
I record the podcast, then AI assists with production, marketing materials, and cover images. After my review and approval, scheduled automations publish and distribute the finished work to the Shared Sapience website, and across my other web and social channels.
More builds
Sam - Agentic Social Media Manager
Desktop AI collaborator that learns your voice from writing samples, generates on-brand multi-platform posts, and auto-schedules via Postiz. Shipped commercially as a lifetime-license product.
Backs my teaching on taking an AI idea from prototype to a product people pay for.
WorkAgent - RAG for Curriculum Work
Document processing and RAG chat for institutional curriculum work: hybrid pgvector search, Docling OCR, and human-in-the-loop review running inside a real college.
Backs my consulting on workplace AI that has to satisfy compliance and real stakeholders.
Obi - Live Podcast AI Participant
An AI room regular on the Open Builder Bar podcast: Discord bot with persistent PostgreSQL memory, plus a live voice loop of Deepgram STT, LLM response, and Cartesia TTS into StreamYard.
Backs my sessions on agent design, memory, and real-time voice pipelines.
Multi-Agent Writing Interface
Web UI wrapping Claude CLI sessions for collaborative long-form drafting, SEO metadata, social responses, and Substack notes, with persistent context per agent.
Backs my teaching on multi-agent workflows for real editorial output.
GPU-Accelerated Transcriber
Speech-to-text web app on Faster-Whisper with real-time VRAM monitoring, model switching from tiny to large-v3, and SRT/TXT export, running on a local RTX 4090.
Backs my workshops on local inference and what consumer hardware can really do.
Video Intake & Publishing UI
Upload interface that archives media, pre-fills YouTube metadata from AI-generated marketing docs, and hands normalized payloads to n8n for the full publishing workflow.
Backs my consulting on workflow automation that removes whole categories of manual work.
The lab underneath it all
Every build above runs on infrastructure I own and operate: a self-hosted stack with full observability, automated backups, and secure networking. When I advise on self-hosting, the advice comes from a system I keep healthy every day.
40+
self-hosted services in one Docker stack
8 nodes
across VLAN-segmented networks with a Tailscale mesh
24/7
monitoring with Prometheus, Grafana, and push alerting
RTX 4090
hand-built Ryzen 9 workstation for local inference
Live observability across hosts, containers, and services
The 40+ service container stack, managed and monitored
AI Image & Video Generation
I work with ComfyUI and Stable Diffusion for AI image and video generation, building node-based workflows for production output. These run locally on the RTX 4090 workstation.
ComfyUI image generation workflow - multi-step node graph with controlnet and upscaling
ComfyUI video generation workflow - segment of a WAN-based video pipeline
AI-generated character art - ComfyUI with LoRA fine-tuning
AI-generated video - WAN model via ComfyUI (5-second loop, 720p)
AI-generated landscape - ComfyUI with custom checkpoint
Want this kind of capability in your organization?
I teach the skills behind these systems and consult on building them. Either way, you work with someone who does this daily.