v2026.07The One Where I Taught an AI to Design LEGO Without Copying ItJuly 2026
✅ Added
- Built a parametric LEGO model generator in Python: a three-skill pipeline (analyzer → architect → generator) that turns a brief into real, validated, buildable LDraw geometry - not a renderer, every part placement is real
- Taught the analyzer to extract design techniques - not geometry - from 7 official LEGO sets (five display helmets plus a 4-helmet F1 system) into a reusable knowledge base of 35 techniques
- Took the Red Ranger helmet build through a full redesign - Revision 2 swapped a boxy Technic-spine scaffold for a thick stacked-plate skull-core, real curved slopes, and an anatomically ordered visor/mouth-grille stack
🔭 Coming Soon
- Open Revision 2 in BrickLink Studio, add part variety, generate real build instructions
- Point the analyzer at more official sets to keep growing the technique knowledge base
🔧 Fixed
- Revision 1's thin spine and floating deck plates - read as a disconnected scaffold - replaced by a skull-core every module attaches to directly
- Generated files that Studio silently mis-resolved now carry the exact per-submodel header format (Name:/Author:/NumOfBricks) Studio expects
⚠️ Known Issues
- The part library's curved-slope and SNOT-bracket entries are a starting subset, not exhaustive - the generator throws on anything unregistered, by design
- The helmet hasn't had its final detail pass in Studio yet
📊 KPIs
7Official sets analyzed
35Techniques extracted
8 modules, 0 errorsRed Ranger helmet
🛠 Skills
Parametric CAD / generative designPython geometry engines (LDraw/LDU)Multi-skill AI pipelines3D collision & connectivity validationReverse-engineering design principles from real products
v2026.06The One Where the Bingo Card Went LiveJune 2026
✅ Added
- Shipped BuzzBoard Bingo end-to-end with Yam Cohen - real-time multiplayer meeting bingo with live rooms, a marketing site, and a post-game analytics dashboard
- Added an Ideas page to Monsets - a public backlog of what's coming next - plus an llms.txt file so AI crawlers read the site correctly too
- Rebuilt this site's build pipeline on vite-ssg for pre-rendered pages and wired up automatic Firebase deploys on every push to main
🔭 Coming Soon
- BuzzBoard Bingo user growth push
- Monsets roadmap shaped by whatever lands in the new Ideas page
🔧 Fixed
- Manual "build then deploy" replaced by a GitHub Actions workflow that ships on push
- BuzzBoard's placeholder branding - a ™ nobody asked for - replaced with a real wordmark and a design-accurate layout
⚠️ Known Issues
- BuzzBoard's mobile layout took three follow-up commits to actually stick
- The Ideas page has zero submitted ideas from real users so far
📊 KPIs
5Days 0 → public launch
80+Commits on BuzzBoard Bingo
2Sites on automatic deploy
🛠 Skills
Real-time multiplayer (Firebase Realtime DB)Product collaboration with a co-founderStatic site generation (vite-ssg)CI/CD (GitHub Actions + Firebase Hosting)Analytics instrumentation (GA4)
v2026.05The One Where the Backend Grew UpMay 2026
✅ Added
- Migrated Monsets' WhatsApp orchestration off n8n onto a real TypeScript backend - message routing, image-generation queueing, payment webhooks, and reminders now run as tested code instead of automation nodes
- Wired Sentry into the new backend so failures page someone instead of getting discovered by an angry parent
- Started Home Helper - an Expo app that tracks attendance, pay, and pension compliance for Israeli households employing domestic workers
- Shipped inventory management with real login for rout's pool-business partners
🔭 Coming Soon
- Home Helper past its one-household beta
- rout inventory live for its first outside partner
🔧 Fixed
- "Ship it and watch prod" replaced by tests written before the code that makes them pass
- Every phase now closes with a code review report and a human sign-off, not just a merge
⚠️ Known Issues
- LaunchCrew v2 is still parked since April - needs unsticking
- Home Helper has exactly one household using it: mine
📊 KPIs
6WhatsApp webhook events shipped
2New codebases started
0P0 ops incidents
🛠 Skills
Backend migration (n8n → TypeScript)Test-driven developmentError monitoring (Sentry)Expo / React NativePhase-based delivery with built-in review gates
v2026.04The One Where Beta Became Real UsersApril 2026
✅ Added
- Synthesized feedback from 10+ LaunchCrew beta users - identified top friction points, prioritized v2 scope
- Shipped this portfolio - designed and built solo with vibe coding
- Formalized an AI design workflow that produces non-generic output: inspiration → AI analysis → execution
🔭 Coming Soon
- LaunchCrew v2 launch with beta feedback fully integrated
- New side project - currently stealth 👀
🔧 Fixed
- "Ship and see" replaced by structured design review before a single line of code
- Async communication gaps - batched and documented instead of reactive pings
⚠️ Known Issues
- LaunchCrew v2 still in iteration
- Occasionally writes a PRD for something that didn't need one
📊 KPIs
10+Beta users managed
2Products live
1 weekTime from feedback → v2 spec
🛠 Skills
User research synthesisProduct iterationAI-assisted design workflowVue 3 / Vite / Tailwind
v2026.03The One Where I Shipped LaunchCrew in 4 Weeks, SoloMarch 2026
✅ Added
- LaunchCrew - a pre-launch audit tool: paste URL, get actionable fixes with AI-ready prompts covering security, SEO, performance, accessibility, and QA
- Defined product scope, virtual team structure (6 personas), and go-to-market - no budget, no team
- 10-person structured beta before any public announcement - feedback loop built into launch process
🔭 Coming Soon
- LaunchCrew v2 with beta feedback integrated
- More experiments on AI + design quality
🔧 Fixed
- Shipping without validation - now always beta before broad release
- Generic AI-generated design - structured visual analysis step added to every project
⚠️ Known Issues
- LaunchCrew needs post-beta iteration before full public launch
- Documentation is functional, not pretty
📊 KPIs
4Weeks 0 → live
10+Beta testers recruited
15+Sites audited in beta
🛠 Skills
Solo product ownership (0→1)Go-to-market strategyBeta program designSEO / security / performance auditingNo-code / vibe coding
v2026.02The One Where I Enrolled in All the Courses and Actually Watched ThemFebruary 2026
✅ Added
🔭 Coming Soon
- Apply Remotion pipeline to Monsets Instagram content
- Turn Maven learnings into shipped features
🔧 Fixed
- Passive learning replaced with applied output - built something with every new skill instead of just taking notes
- Video production treated as a manual task - automated the pipeline with code
⚠️ Known Issues
- Maven wishlist is longer than available hours
- Day 30 videos look significantly better than day 1
📊 KPIs
30Videos shipped
30Days of consistency
ManyMaven courses enrolled
🛠 Skills
Remotion (programmatic video)React-based video templatingContent automationContinuous learning - Maven
v2026.01The One Where a WhatsApp Bot Made Parents Cry (Happy Tears)January 2026
✅ Added
- Monsets - WhatsApp bot that generates themed monthly milestone cards from a baby photo. Parents pick a theme, bot handles the rest
- Automated Instagram content pipeline - every reel on @monsets.co generated via Claude Code, zero manual editing
- Full marketing site built end-to-end with vibe coding - no designer, no dev team
🔭 Coming Soon
- Expanded theme library based on user requests
- More automation in the image generation pipeline
🔧 Fixed
- "A marketing site needs a designer" - proved otherwise
- Solo development isolation - shipped with real parent feedback from day one
⚠️ Known Issues
- WhatsApp Business API has strong opinions. Respected, still annoying.
- The catalog page filter UX needs a proper pass
📊 KPIs
12+Milestone themes shipped
20+IG reels auto-generated
~14Days concept → live product
🛠 Skills
WhatsApp Business APIAI image generationEnd-to-end product developmentContent automationClaude Code / vibe coding