The need: Photography shot with a fully manual anamorphic lens produces squeezed images. Every frame needs de-morphing before editing. After research — nothing existed to batch-process this.

What I built: macOS desktop app. Electron + React + TypeScript. Uses dnglab for RAW conversion, ffmpeg for de-squeeze, exiftool for metadata. Five modules. Drop files in, get corrected images out.

Why it matters: First time I said “Hello Claude, build me something” — and it worked. A real tool solving a real problem. The spark that lit everything that followed.

→ GitHub: wolfschram/anamorphic-lens

The need: An AI second brain. Something that captures what I say, understands what matters, and helps me remember what my ADHD brain forgets. Nothing on the market was designed for neurodivergent minds.

What I built: Three versions. v1: concept docs + MCP server spec. v2: Swift/Xcode iOS audio capture. v3: Electron desktop app with SQLite brain.db. Each closer than the last.

What failed: All of it. None became a usable product. The vision was right but the execution kept hitting walls I didn’t know existed yet.

What I learned: Failure is the fastest teacher if you’re honest about it. Most people quit at attempt three. The breakthrough comes after you stop blaming the tools and start understanding the problem.

→ jenny-app · capture · assistant

The need: Same as Jenny — but this time I understood the problem well enough to solve it properly.

What I built: Full desktop app. Swift audio capture as separate process. Real-time transcription. Claude API for intelligent extraction of tasks, promises, memories. Supabase cloud backend. Post-it style overlay — max 15 words — surfacing context exactly when you need it.

Why it works: ADHD-first design as the core principle. Every feature asks: does this reduce cognitive load or add to it? If it adds, it doesn’t ship.

What it proved: 20 failures lead somewhere if you don’t quit. First tool built entirely with AI assistance that works in daily use.

→ GitHub: wolfschram/lighthaus-v1

The need: Lighthaus worked as a personal tool. Could it become a product? If I can’t sell the app, maybe I can sell myself to sell the app. Nobody was building AI assistants focused on neurodivergent people.

What I built: The Intelligence Gap — strategic vision mapping the gap between AI capability and human understanding. The Intelligence Layer — deep technical architecture with 13 microservices, hybrid processing, $16/user cost model.

What happened: Neither led to a commercial outcome. Reached out to CEOs on LinkedIn — heard nothing. The market is overloaded with people having their second spring with AI. But the concept is still valid.

What I learned: Narrative alone doesn’t sell. Architecture alone doesn’t sell. What sells is proof of work and the ability to adapt when the world moves faster than your plans.

The need: 25 years of photography from 55+ countries. ~1,000 curated images sitting on hard drives. Time to bring them to the world as fine art.

What I built: React frontend on Cloudflare Pages. Python FastAPI agent. Electron + React Studio desktop app. Node.js mockup service. Stripe payments. Pictorem print fulfilment. Two-machine workflow — MacBook Pro as creative workstation, Mac mini as render server.

Why it matters: First truly complex multi-system build. Three services sharing data — change one, test all three. Live with real customers and real payments. Not a prototype.

What it taught: How to build properly. How systems depend on each other. How to manage complexity across codebases. Everything after was faster because of what this demanded.

The need: The website alone isn’t enough. Managing a photography business means content ingestion, mockup generation, social media posting, sales tracking, gallery submissions, website control, and AI-powered content creation. All from one place.

What I built: A full desktop application with 29 pages organised into 6 clusters: Content management (import, gallery, sync), Mockups (preview, templates, batch processing), Sales (Pictorem, Artelo, promo codes, licensing), Social & Marketing (Instagram, Pinterest, Etsy integrations), an AI Agent system (dashboard, compose, content queue, pipeline monitor, health panel, CaFÉ export), and Website Control (live control panel, about editor, analytics).

The infrastructure: Three interconnected services that share data. The Electron Studio app on port 3001. A Python FastAPI agent on port 8035 handling Instagram, Pinterest, Etsy, Google Sheets, and Telegram integrations. A Node.js mockup service on port 8036 for template rendering. The three-system rule: change one, test all three.

Why it matters: This is the most complex thing I’ve built. 29 pages, 3 services, 6 platform integrations, all communicating. The architecture came from my engineering brain — in the macro world it works one way, in the micro world of code, the same principles apply.

The need: A major broadcast manufacturer needed competitive intelligence across the entire IP routing vendor landscape — IPMX, SMPTE 2110, where the industry is heading.

What I built: Full competitive analysis platform. Research across Grass Valley, Evertz, Lawo, EVS, Blackmagic, Cobalt, Imagine and more. Interactive visualisations. Vendor comparison matrices. Market positioning analysis.

Why it matters: Proved the intersection of domain expertise and AI-augmented building. I sat with them, had done my homework, and could answer questions that came easy because of 25 years in the industry. Built in days, not months.

→ GitHub: wolfschram/riedel-intelligence

The need: Job searching is tedious and draining — especially with ADHD. Automate the pipeline: Claude scores listings, researches companies, drafts cover letters. A bot submits. A dashboard tracks everything.

What I built: Node.js server. SQLite with 18 tables. 13-tab command center. Gmail OAuth. Claude API for scoring and cover letters. MCP server with 9 tools. Auto-start daemon.

What failed: The ATS bot has never been tested end-to-end on a real application. Every platform actively fights automation — and they’re better at it than me. Ambitious but fragile.

What I learned: Automation without reliability is just complexity. Building something that works sometimes is worse than building something simple that works always.

The need: After all the failures and successes, pivots and breakthroughs — what do you do with everything you’ve learned?

The answer: You teach it. You share it. You help others make the same transformation. Not from a whitepaper — from lived experience. From doing it, failing at it, succeeding at it, and documenting all of it.

Culture architecture informed by 25 years of leadership and months of intensive AI-augmented building. The Restless Mind doesn’t stop. There will be a Project 09.