Skip to main content

Our Work

A running log of things we've shipped (and killed)

Real projects. Real results. Real talk about what didn't work.

OPS

Southern Turf Carolinas

for Southern Turf Carolinas

Full-stack rebuild for commercial turf installer. Turned 8-second WordPress nightmare into sub-second static site with smart lead capture.

Results:

  • Page load: 8 seconds → 0.7 seconds
  • Mobile bounce rate: 87% → 31%
  • Lead quality up 3x with smart forms
  • First page Google for key terms
AstroVercelTypeScriptResend
OPS

Analytics Pipeline Overhaul

for Stealth E-commerce Brand

Rebuilt their entire data pipeline from Segment chaos to clean, reliable BigQuery warehouse with actual business metrics.

Results:

  • Data freshness: 24 hours → 15 minutes
  • Pipeline failures: Daily → Monthly
  • Analytics trust: 'Nobody believes these numbers' → 'Finally, real data'
  • Cost: -60% despite 10x more data
BigQueryDataflowdbtPython
LABS

Quick & Dirty Market Test

for B2B SaaS Startup

Two-week sprint to test if anyone would actually pay for their 'revolutionary' idea. Spoiler: they wouldn't.

Results:

  • Built functional MVP in 10 days
  • Tested with 50 target users
  • Conversion rate: 0%
  • Saved them $2M in wasted development
Next.jsSupabaseStripe
FALSE START

Animated Brand System

for Creative Agency

Started building a generative logo system. Got distracted by cooler project. Will return someday.

What happened:

  • Built working prototype
  • Created 3 variations
  • Got distracted
  • ...
ReactFramer MotionTypeScript
FALSE START

The Perfect CMS

for Internal

Every developer's white whale. Started building the CMS to end all CMSs. Realized Notion exists.

Why it's paused: Someone else did it better

What happened:

  • Beautiful schema design
  • 2 weeks of work
  • Discovered Payload CMS
  • Deleted everything
Next.jsPrismatRPC
FALSE START

Real-time Collaboration Tool

for Internal Project

Building a Figma-like multiplayer editor. Hit WebRTC limitations. On pause while we learn more.

Why it's paused: Underestimated complexity

What happened:

  • Built working prototype with CRDT
  • Sync worked... mostly
  • Discovered why Figma has 100 engineers
WebRTCCRDTTypeScriptCanvas API
FALSE START

ML-Powered Content Generation

for Content Startup

Auto-generate blog posts from bullet points. GPT-4 came out mid-project. Pivoted to using their API.

Why it's paused: OpenAI did it better

What happened:

  • Trained custom model on 10k articles
  • Achieved 70% quality score
  • GPT-4 scored 95%
  • Client happy with API solution
PythonTensorFlowGPT-4 APIFastAPI
GRAVEYARD

ML-Powered Sentiment Dashboard

for Marketing Agency

Client wanted AI sentiment analysis. Built a prototype. Realized Google Trends did 90% of it for free.

Cause of Death: Death by existing Google Trends

What we learned:

  • Built working prototype in 1 week
  • Discovered free alternative
  • Saved client $200k/year
  • Killed with prejudice
PythonTensorFlowBigQuery
GRAVEYARD

Real-Time Inventory Sync

for E-commerce Platform

Everyone said they needed real-time. Tested with 15-minute batches. Nobody noticed.

Cause of Death: Murdered by batch processing

What we learned:

  • Built real-time prototype
  • Tested batch alternative
  • Users couldn't tell difference
  • Saved 90% complexity
KafkaRedisPostgreSQL
GRAVEYARD

Blockchain Loyalty Points

for VC-Backed Startup

They wanted blockchain everything. Built both versions. Postgres was 100x faster and free.

Cause of Death: Postgres did it better

What we learned:

  • Blockchain version: 3 second transactions
  • Postgres version: 30ms transactions
  • Cost difference: $50k vs $500/year
  • They picked Postgres
SolidityWeb3.jsPostgreSQL
GRAVEYARD

Microservices Migration

for Mid-size SaaS

Client wanted 47 microservices. Peak traffic was 12 requests/second. Kept the monolith.

Cause of Death: Monolith was fine

What we learned:

  • Analyzed actual traffic patterns
  • Peak load: 12 req/sec
  • Microservices needed: 0
  • Money saved: $500k
DockerKubernetesRails