Painting Direct
A binding paint quote, straight from the painter to the homeowner. No contractor SaaS. No lead-gen middleman. No upsell. Ninety seconds, real price.
Binding estimate
$4,820
740 sf · 4 doors · valid 14 days
LaborMaterialsOverhead+ Premium color
The trades run on middleman SaaS — Bolster, JobNimbus, ServiceTitan — built for the contractor, not the customer. Lead-gen sites pretend to give homeowners "honest quotes" and then sell the lead to the highest bidder. Painting Direct cuts all of that out: a small painting business gives a homeowner a binding quote in ninety seconds, with nobody in between. The painter edits their own rates with a full audit trail. The homeowner gets a real price, signs in two taps, and the work begins — same relationship as walking into a neighborhood shop, with software that respects both sides.
- Audience
- Small business ↔ homeowner · nothing in between
- Pattern
- Trade-specific direct-to-customer estimator. Painting today; the same shell fits a plumber, an electrician, a window installer.
- Stack
- Next.js 16 · Cloudflare Pages · Lambda · DynamoDB · S3 · SES
- Status
- v0.1 live · roadmapping AR room measurement
02
• Private · family pilot
Custom Job Searches
A specialty-aware healthcare job board with a 5:30 AM daily podcast that reads new listings out loud over coffee.
cron0 5,11,17,23 * * *
scrapersWorkday · Phenom · DocCafe · Sitemaps
filterPA · POCUS ✓ · new-grad ✓
05:30 ET→ Polly (Joanna) → inbox.mp3
Built for a graduating Physician Assistant who needed specialty-aware listings the generic boards don't provide — PA separated from physician, POCUS tagging, new-grad-welcome flags. Every six hours a scraper fleet hits eight hospital systems. A custom AI watchdog self-heals scraper drift and only escalates when it can't fix itself.
- Audience
- Family pilot → vertical SaaS candidate
- Stack
- Lambda · DynamoDB · SES · Polly · custom AI scrape & QA
- Status
- Live since 2026-05-13 · 141 listings at launch
Faces
A lifetime of family photos in one place — from every provider, every device. You're in charge: faces, scenes, places, dates. Find what you love. Scrub what you don't. No vendor lock-in. No corporate bullshit.
find_and_delete("ex_boyfriend")
247 photos · 12 albums · one tap
Your photos shouldn't be hostage. They accumulate across iCloud, Google Photos, Amazon Photos, old phones, dead hard drives — and each vendor keeps them, indexes them, monetizes them, and never quite lets you leave. Faces pulls everything into one library you own, indexes faces and scenes and places, and lets you actually do what you'd expect: every photo of Mom in Hawaii in 2008, every photo of the kids' birthdays in one timeline, every photo of an ex you'd rather not see again — find them, organize them, delete them. Tag what matters. Scrub what doesn't.
- Audience
- Anyone with a decade of photos and zero control over them
- Ingests
- iCloud · Google Photos · Amazon Photos · local drives · phones
- Stack
- AWS Rekognition · Lambda · S3 · Cloudflare Pages
- Status
- Production pipeline running over a 71k personal archive · consumer version in development
hiebel.ai/faces · demo coming
04
• Partner build · prototype
Capere
Always-on cameras that stream end-to-end encrypted into a vault only the owner can open, with retroactive "capture the moment" buffering. Someone else's product vision; we brought it to life.
Recording · E2E · XChaCha20
−00:90 ← now
save
The doorbell rang → save the 90s before now.
The product idea isn't ours — it came from a design partner who'd been thinking about cameras that don't leak everything to a cloud vendor, with the ability to grab the last thirty seconds of something interesting after the fact. We built it. Prototype landing and working dashboard mock shipped first so the partner could react to feel and trust before any hardware or streaming work began. Two form factors — Orb and Sentinel. Optional private mode keeps everything on-prem; the architecture scales identically from a single home to an enterprise campus. This is something we do: take someone else's product vision from sketch to shipping prototype.
- Audience
- Privacy-skeptical home → enterprise on-prem deployment
- Origin
- Partner concept · we built it
- Stack
- E2E encryption · ring buffer · vault appliance · on-prem option
- Status
- MVP prototype shipped 2026-04-13
05
• Alpha · Landing public
RESQD
A quantum-secured digital vault for the things you want to survive you — with cryptographic proof that nobody has peeked.
CipherXChaCha20-Poly1305
WrapML-KEM-768 · post-quantum
HashBLAKE3
ShardsReed-Solomon 4+2 · S3 + GCS
AnchorBase L2 · ResqdCanaryAnchor.sol
•—•—•—•—•—•—°
Canary intact · access count: 6
Files are encrypted client-side in WebAssembly, split by erasure coding across S3 and GCS, and every access rotates a canary anchored on Base L2. If anyone — operator or AWS — touched it, the canary count is off and the owner knows. Same architecture serves a single family heirloom and an enterprise-scale audit trail.
- Audience
- Estate-planning consumer → enterprise compliance vault
- Stack
- Rust → WASM · Lambda ARM64 · S3 + GCS · Solidity / Base · Next.js 16
- Status
- Consumer alpha live (April 2026) · invite-only
06
• Live · sample on request
Auto Deep Dive
An honest car buyer's playbook for any listing URL. The version of CarGurus and TrueCar that doesn't sell your contact info to a dealer.
Inputany AutoTrader / cars.com / CarGurus / CarMax URL
Market valueKBB · Edmunds · Carvana · comp set
Risk auditNHTSA recalls · reliability flags · dealer reputation
PPI checklistyear-/model-specific inspection points
Offer ladderopening → walk-away · state-specific gotchas
URL → research → PDF → inbox
Typical turnaround: ~3 minutes · no lead-gen, no upsell, no tracking
Every "honest car research" site on the internet is a dealer lead-gen funnel. This one isn't. Drop a listing URL in; get back a multi-page PDF playbook covering market value, dealer reputation, reliability risks, a year-/model-specific PPI checklist, an offer ladder, the exact scripts to use at the dealer, and the state-specific tax and title gotchas. Powered by custom AI models that run the research end-to-end at near-zero marginal cost — which is what lets us skip the data-resale model that funds the competition.
- Audience
- Family car buyer → fleet purchasing research
- Stack
- FastAPI · custom AI research orchestration · WeasyPrint · SES · Cloudflare tunnel
- Status
- Live · private beta · public landing & sample PDF coming