Building for
privacy by default.

Cloud coding assistants cost money every month and route your source code through someone else's servers. LocalVocal removes both barriers — a complete interactive guide to run a private, $0 AI coding assistant on a 6 GB NVIDIA laptop GPU in about fifteen minutes.

• Developers who want Copilot-style help without a subscription
• Anyone with sensitive or proprietary code
• People on modest GPUs (6–8 GB) who assume local AI is out of reach
• Tinkerers who want to own their entire stack

Privacy by default and zero-cost by design. The single-file HTML delivery is itself a statement: no frameworks, no npm install, no cloud dependency. Local RAG with nomic-embed-text means cloud calls send a few hundred tokens, not whole files.

• Embedded VRAM budget visualizer shows GPU headroom in real time
• Gemini panel pre-primed with your exact hardware context — off-script questions get grounded answers
• Bridges the "AI is only for expensive hardware" perception gap
• Doubles as a reusable template for any GPU target

Most users are unaware of ghost apps (uninstalled apps that leave residual data and permissions) and stale trackers still exposing personal information. A plain-English digital hygiene guide — raising awareness and directing action, without any server-side irony.

• Non-technical smartphone users who want to understand data exposure
• Privacy-conscious users who suspect apps do more than they should
• Parents and educators looking for digestible mobile hygiene resources
• Anyone who uninstalled an app but never thought about leftover data

Education before action. The site is deliberately static with no server-side tracking — no analytics backdoor, no irony. The modular form provider approach demonstrates progressive enhancement: start free, swap providers as needs grow.

• Ghost app prevention is an underexplored consumer problem — first-mover awareness value
• Designed to double as a lead-gen funnel — product thinking in an educational wrapper
• Natural companion to AuditAd: Know-ur-phone educates, AuditAd acts
• Practices what it preaches: static site with no tracking teaches privacy without hypocrisy

Android apps engage in a hidden economy of cross-app profiling and background data harvesting. Existing solutions route traffic through third-party VPN servers (a new privacy risk) or require root access. AuditAd uses a local DNS loopback on Port 53 — entirely on-device, nothing sent externally.

• Privacy-conscious Android users wanting visibility into app telemetry
• Security researchers and mobile developers auditing permissions
• Developers who want to audit their own manifest before publishing
• Users who want to neutralise cross-app profiling without rooting

Zero-knowledge, on-device. The Bloom Filter checks 150,000+ domains in sub-millisecond time, entirely in-kernel. "Trust but verify" becomes "verify, then trust." The interactive Bloom Filter Playground is a transparency mechanism that shows users exactly how the matching engine works.

• Bloom Filter Playground: interactive CS education inside a consumer privacy app — tune m, n, k parameters and see real false-positive probabilities
• 64-bit memory grid visualises how trackers "flip bits" — abstract algorithm made visceral
• Package Auditor with Gemini AI risk scoring — developer feature tucked inside a consumer app
• Most technically ambitious: Capacitor Java bridge, custom AppCheckPlugin, Bloom Filter, and Gemini AI all in concert

AI prompts are treated as throwaway one-liners rather than reusable professional assets. Without structure, teams rewrite the same prompts from scratch, get inconsistent AI behaviour, and lose institutional knowledge. PromptOS frames prompts as structured artefacts: skills, rules, and boundaries — versioned in a repo.

• Developers and AI engineers who want to manage prompts as code
• Teams building LLM-powered products needing consistent AI behaviour
• Prompt engineers tired of reinventing the wheel each project
• Agent builders who want a structured skill/boundary framework

Prompts as first-class software artefacts. The skills/rules/boundaries taxonomy mirrors AI alignment thinking — capabilities, constraints, guardrails. Encoded in files in a repo, it fits any workflow without a new platform. "If it lives in your repo, it lives in your workflow."

• The .ai/ folder convention is an elegant, unopinionated primitive — any editor, CI, or LLM can consume it
• The taxonomy maps directly to how modern LLM system prompts work — drew from AI safety alignment thinking
• Earliest repo chronologically — the philosophical seed that spawned every other project
• Significant potential as a VS Code extension or CLI tool — groundwork already there

Most AI writing tools are cloud-dependent, per-token-billed, and treat content generation as a single-shot prompt. MegaMind is built around a production pipeline: research, draft, reformat for a specific platform, export — entirely local, two modes: interactive Electron UI or headless folder-watcher service.

• Content creators, LinkedIn influencers, and solopreneurs with high content volume
• Developers who want a local AI writing pipeline without per-token costs
• Storytellers and comic-narrative creators (first-class use case)
• Windows power users who prefer folder-based automation over SaaS dashboards

Content creation as a production process, not a conversation. Clean split between generation and formatting stages — separation of concerns, explicit API contracts (FastAPI schemas), observable state. The folder-service contract echoes Unix pipe philosophy: drop a job in, pick up a finished artefact.

• Dual-mode architecture respects two very different human workflows: creative (UI) and automated (headless)
• The 'storyteller' default model and comic-style job examples reveal a creative writing ambition beyond business content
• worthifyme.com domain suggests a larger brand vision — the studio is an engine for a broader creator economy play
• Most production-ready: published release v1, CHECKSUMS.txt, proper release notes — mature distribution thinking
• Folder-service contract mirrors enterprise ETL pipeline patterns — Pentaho/PDI thinking leaking into indie software