Home

Case Studies

Real transformations. Real results. Actually deployed. Actually running.

FEATURED$200K CAPTURED

WOTC Tax Credit Transformation

The Context

WOTC (Work Opportunity Tax Credit) allows employers to claim $2,400–$9,600 per eligible hire. It's a federal program designed to incentivize hiring from disadvantaged groups.

The problem: Most employers leave massive money on the table.

The Problem

Question 13 on the WOTC form asks whether the employee was unemployed, wanted more paid work, or felt underemployed when they started the job.

Here's what happens:

  • ×Employee sees confusing legal language
  • ×Doesn't understand what it's actually asking
  • ×Clicks “No” to be safe
  • ×But they actually qualify
  • Employer loses $5,000+ in tax credits

Multiply this by hundreds of employees. You're looking at six figures in lost tax credits annually.

The industry had been trying to solve this for years. Better form design. Clearer instructions. Training programs. Nothing worked. Wrong answers remained stubbornly high: 30%.

The Insight

“Applicants weren't clicking 'No' because they weren't eligible. They were clicking 'No' because they didn't understand the question.”

This wasn't a UI problem. It wasn't a training problem. It was a comprehension problem.

The bottleneck wasn't technical—it was human understanding.

The Solution

I built an audio-guided form.

Audio Guidance
Questions read aloud
Simple Yes/No
Big buttons, mobile-friendly
60 Seconds
Complete verification

Every question is read aloud in plain language. The employee hears what the question actually means. They answer correctly. The tax credit is captured.

The Results

MetricLegacy MethodsAudio Form
Wrong answers30%< 5%
Credits LOST$5,000+ eachNearly zero
Completion time45 minutes60 seconds
Build time1 week
Credits CAPTUREDMany missed$200K+

One person. One week. $200K in tax credits captured that legacy methods missed.

The insight wasn't “build a better app.” The insight was “the bottleneck is comprehension, not code.”

Timeline

2024-06First mention of WOTC(Problem identified)
2024-08Comprehension insight crystallizes(Root cause found)
2024-08"Simple > Smart" principle emerges(Solution approach)
2024-09Audio WOTC built(1 week, solo)
2024-12$200K revenue generated(Production success)
View full WOTC case study
FEATUREDCOGNITIVE PROSTHETIC

Brain MCP — Cognitive Prosthetic

The Context

Asperger's + ADHD = monotropic attention. One tunnel at a time. Total immersion — and total amnesia for everything outside it.

Every context switch was a factory reset. Decisions remade. Questions re-asked. Breakthroughs re-discovered months later.

The Problem

Context dies when the tunnel moves. Three months deep in Torah study? The AI architecture decisions don't exist anymore. Switch to frontend? The 47 open Torah questions vanish. Not deprioritized. Gone.

This isn't poor organization. It's neurology. Monotropic attention doesn't do “background threads.”

The Insight

“The Bottleneck IS the Amplifier. Don't fix monotropic attention — build infrastructure that preserves context across attention shifts.”

Log everything. Embed it. Make it queryable. Let the tunnel do what it does best. Give it a safety net.

The Solution: 25 MCP Tools

377K messages indexed across Claude, ChatGPT, Claude Code, and Clawdbot. 82K semantic embeddings for conceptual search. 9,979 structured summaries with extracted decisions, questions, and breakthroughs.

tunnel_state
Reconstruct save-state for any cognitive domain. Load a saved game in 12ms.
context_recovery
Full re-entry brief: thinking stage, open questions, last decisions.
switching_cost
Quantified cost of moving attention between domains.
trust_dashboard
System-wide proof the safety net works. Coverage, sync health, freshness.

+ 4 more prosthetic tools (open_threads, dormant_contexts, cognitive_patterns, tunnel_history) and 17 generic tools (semantic search, thinking trajectory, alignment check, etc.)

The Results

MetricBeforeWith Brain MCP
Past contextGone when tunnel moves256ms retrieval
Open questionsForgotten permanently111,942 preserved
DecisionsRemade from scratch36,743 tracked
Domain switchingFull restartCost quantified + brief
Context survival0%100%

One person. 18 months. A cognitive prosthetic that turns neurological context death into queryable institutional memory.

The insight wasn't “fix the attention system.” The insight was “the bottleneck IS the amplifier — build a safety net around it.”

Timeline

2023-06First AI conversations begin(Raw data accumulates)
2024-08Context death problem crystallizes(Pattern recognized)
2025-02Parquet pipeline built(Messages become queryable)
2025-06LanceDB vector search added(Semantic search enabled)
2025-11Cognitive prosthetic tools built(8 tools for monotropic attention)
2026-0225 MCP tools, 377K messages, 143/143 tests(Production-grade)

Technical Stack

LanceDB + DuckDB
Vectors + Structured data
nomic-embed-text-v1.5
768d embeddings, local
Model Context Protocol
25 tools, any client
View full Brain MCP case study

More Case Studies

View all 132 repos

Have a bottleneck that needs finding?

48 hours to identify it. 5 days to prove the solution. Free until value is demonstrated.

Let's Talk