Apps Portfolio Report

A full audit of every project repo inside ~/Desktop/apps

Generated Saturday, 23 May 2026 51 projects analysed ~62 GB on disk node_modules & raw data dirs skipped

01Executive Summary

This report inventories every project folder in your apps directory — what each one is, how it is built, how healthy it looks, and what needs attention. Think of it as a stock-take of a very busy workshop: lots of half-built machines, a few finished ones, and some sawdust to sweep up.

51
Project folders (excl. node_modules)
~62 GB
Total disk footprint
~39 GB
Is ONE raw-data folder (EPC Hunt CSVs)
16
Projects with NO version control
11
Projects flagged for exposed secrets

The big picture. This is a prolific solo-builder portfolio — 51 distinct projects, almost all AI-powered products, built fast and iterated hard. The dominant pattern is unmistakable: TypeScript everywhere, Cloudflare and Vercel for hosting, and an AI model (Gemini, Veo, fal.ai, Claude, or the Vercel AI SDK) wired into nearly every app.

Headline observations. (1) One folder, ai-retrofit-radar, is 40 GB — and 39 GB of that is a raw CSV scrape sitting inside the project. (2) The same real-estate AI-video product exists as roughly five near-identical copies. (3) 16 projects have no git at all — including multi-gigabyte ones — so there is no history or backup. (4) Several repos carry committed .env files; secrets should be rotated. (5) A handful of repos have 100+ uncommitted changes — unsaved work at risk.

Maturity. A small core is genuinely substantial (paperclip, ai-retrofit-radar, ai-video-listing, cfai-frontend). The majority are work-in-progress or have gone quiet. Around a dozen are pure artifacts — research write-ups, planning folders, wikis — not software at all.

  PORTFOLIO SNAPSHOT
  ------------------------------------------------
  Projects analysed      51      (node_modules skipped)
  Total on disk          ~62 GB  (~39 GB = 1 data folder)
  Real project code      ~22 GB
  Under version control  35 / 51
  No git at all          16 / 51
  Stale > 4 months       ~14
  Secret-exposure flags  11 projects
  Biggest single repo    ai-retrofit-radar (40G)
  Most commits           cfai-frontend (839)
  ------------------------------------------------
  Stack DNA: TypeScript-first . Cloudflare-heavy . AI-native

02Aggregate Dashboard

The same 51 projects, broken down three ways. Bars are counts of projects.

By apparent status

By primary language

By size on disk

Themes that jump out

  • AI real-estate video (x6)ai-video-listing, ai-video-platform, lux_home_ads, lux_home_ads-autoimprove, ai-subtitle-platform and ai-tribute-video are all the same engine. Huge consolidation opportunity.
  • Crypto & trading (x5) — Alpha Arena (algo_app/algo_be), CoinFantasy (cfai-frontend/coin-fantasy-be) and a Polymarket bot.
  • AI agents & orchestration (x5) — led by paperclip, plus Claude Agent SDK samples and an AI marketing-agency crew.
  • Construction / inspection compliance (x4) — BuildCheck / Tower Shield appears twice.
  • Clones & reverse-engineering — Granola, Claw, Alpha Arena, GetScope all cloned.
  THE TYPICAL PROJECT IN THIS FOLDER

   +------------+    +--------------+    +-------------+
   |  Frontend  |-->|   API layer  |-->|   Storage   |
   | Next.js or |    | Hono on CF   |    | D1 / Convex |
   | TanStack   |    | Workers      |    | R2 / Postgres|
   +------------+    +--------------+    +-------------+
         |                  |                  |
         +-- AI: Gemini . Veo . fal.ai . Claude --+

  Hosting:  Cloudflare (Workers/D1/R2) + Vercel
  Auth:     Better Auth / Clerk     Payments: Polar
  Tooling:  pnpm + Turborepo . oxlint . Playwright

03Portfolio Table

All 51 projects. Click a column header to sort; use the buttons and search box to filter. Click a project name to jump to its detailed card.

Project Category Language Status Size Last commit Description

04Project-by-Project Detail

Every project in full, ordered largest-first. Each card shows what it is, how it is built, its git health, and anything notable or risky.

05Flags & Recommendations

The things worth acting on, grouped by urgency. These are the cleanup and risk items the audit surfaced across all 51 projects.

Disk The 40 GB problem

  • ai-retrofit-radar is 40 GB — about 39 GB is a raw CSV scrape under data/dedup/. That single folder is roughly two-thirds of your entire apps directory. Raw data should not live inside a code repo: move it to an external drive, an R2 bucket, or a dedicated data lake, and add data/ to .gitignore.
  • Other multi-GB folders (ai-tribute-video, Autoshoot-AI, talk_to_my_newsletters, ai-subtitle-platform, propgenie_backend) are large mostly because of node_modules, .git history and build output — not source. A periodic node_modules sweep on dormant repos would reclaim a lot.
  • A stray top-level node_modules (178 MB) plus a root package.json / package-lock.json sit directly in apps/. That is not a project — delete it.

Risk Exposed secrets & no version control

  • Committed .env files in admin-site, medshoot-ai, meal_planner_be, propgenie_backend and propgenie_front_end; committed .env.production / .env.staging / .env.development in ai-retrofit-radar, ai-subtitle-platform, ai-tribute-video and aimenu. Action: rotate any live keys, then git rm --cached the files and add them to .gitignore.
  • aifr_webapp keeps .env and .env.local as plain files and has no git — secrets are both unprotected and unversioned.
  • 16 projects have no git repository at all: Autoshoot-AI (2.3 GB!), ai_inspection_app, ai_employee, aifr_webapp, buildcheck, granola-clone, digital-marketing-agency, inner-circle-graph-rag, cloudflare-cookbook, sam-app, hermes-crash-course, inner-circle-build, exp, perper_clip_testing, sprintplan-wiki, static_site_generation. At minimum, git init && git add -A && git commit the real apps among them.
  • inner-circle-graph-rag hard-codes a Neo4j password in its README.

Warn Unsaved work — commit before it is lost

  • Large uncommitted working trees: aimenu (230 changed files vs only 6 commits), ai_build_check (115, and no branch checked out), claw_clone (61), claude_agent_with_aws_agent_core (50), content_generato_mission_control (31), medshoot-ai (22), ai-video-listing (20), talk_to_my_newsletters (20), frenchico (19).
  • frenchico has a single "Initial commit from Create Next App" and 19 uncommitted files — essentially nothing is saved.
  • ~20 folders have no README.md. The real apps among them (ai_inspection_app, ai_build_check, claws_tracking, content_generato_mission_control) would benefit from a one-paragraph README so future-you knows what they are.

Consolidate Heavy duplication

  • One product, five-plus copies. The luxury real-estate AI-video SaaS exists as ai-video-listing, ai-video-platform, ai-subtitle-platform, lux_home_ads and lux_home_ads-autoimprove — same architecture, same domains, even an identical copied README. Pick one canonical repo; archive the rest into an _archive/ folder.
  • BuildCheck / Tower Shield exists twiceai_build_check and buildcheck. AI fashion photography exists twiceAutoshoot-AI and medshoot-ai. sample-claude-code-web-agent-on-bedrock-agentcore exists both as its own folder and nested inside ai_employee.

Archive Stale & abandoned candidates

  • Untouched longest: algo_be (Oct 2025), sample-claude-code-web-agent... (Nov 2025), propgenie_backend & propgenie_front_end (Jul 2025 — ~10 months).
  • Quiet since Dec 2025: coin-fantasy-be, cfai-frontend, claude_agent_with_aws_agent_core, algo_app, admin-site, meal_planner_be.
  • inner-circle-graph-rag calls itself a "throwaway prototype" yet occupies 521 MB — harvest learnings and delete.

Keep The healthy core

  • paperclip — most mature: MIT-licensed, 766 commits, full CI/CD, real OSS project.
  • ai-retrofit-radar — most actively developed of your own products (once the data dump is moved out).
  • hemplane — the tidiest small repo: clear README, zero uncommitted changes.
  • Suggested top-level tidy: create _archive/, _reference/ and _artifacts/ folders and sort the dormant, learning, and non-code items into them so the active products are easy to find.