Home

From Zero to 500K+ Lines

The AI-Powered Programming Journey

Started November 30, 2022 with ChatGPT 3.5. Zero coding experience. Now: 523,847 lines, 11 production systems.

The Numbers

523K

Lines written

1,059

Python scripts

343

TypeScript files

11

Production systems

The Beginning: November 30, 2022

"It is amazing to think I wrote my first code with ChatGPT 3.5 from Nov 2022 till today—almost 3 years. I just dived head first into AI starting with ChatGPT 3.5 on 30 Nov 2022. Before that I had never coded."

Day one: a simple Python calculator. The first working function felt like magic.

(Note: I had built no-code apps with AppSheet during the optics decade—but never written actual code. The distinction matters: no-code tools hide the programming; ChatGPT taught me to write it.)

The Before

1999-2003 — Bachelor of Business Science in Finance & Economics, University of Cape Town. Top 1% South African qualification.
2004 — Came to Israel at 24 "for 3 months." Stayed 20 years.
2004-2012 — Almost a decade learning Torah full-time in kollel. First principles thinking. Systematic analysis. Deep pattern recognition.
2012-2022 — The Optics Decade. Built Israel's 4th freeform lens lab. Raised $3M. Self-taught manufacturing at 3 nanometer tolerance. ISO 7 cleanroom.
Nov 30, 2022 — ChatGPT 3.5 released. Everything changed.

The Timeline

Month 1-3: First Steps

Simple Python calculator. First API call (OpenAI GPT-3.5). First web scraper with BeautifulSoup. Learning Python basics, API concepts, JSON handling.

Month 4-6: Building Confidence

First Flask web app. PostgreSQL basics. Email processing automation. Skills gained: SQL, HTML/CSS, Git.

Month 7-9: Scaling Up

First production system: tax credit eligibility checker. First React SPA. ML classification system. Breakthrough: 100+ Python scripts.

Month 10-12: Enterprise Solutions

WOTCFY Platform launch. Processing 10K+ documents. GPT-4, embeddings, fine-tuning. First $100K in tax credits processed.

Year 2: Exponential Growth

YouTube Pipeline: 32,579 videos processed. Chat Analysis: 11,494 conversations. ML accuracy: 94%. Code volume: 400K+ lines.

Year 3: Production Mastery

Multiple enterprise deployments. 500K+ total lines. Claude Sonnet, GPT-4 mastery. $200K+ revenue generated.

The Cognitive Toolkit

Sniper, Not Machine Gunner

"I am a better sniper than machine gunner. Everything I do is like a sniper with a rifle. That is how I think."

Depth Processor

"I am a depth processor not frequency. Go deep into complex problems. Analyze from multiple angles, see patterns others miss."
"Core skills are being able to simplify complex domains and shape solutions tailored for specific situations."

The Learning Methodology

AI as Teacher, Not Crutch:

1. Ask AI to explain concept

2. Write code myself

3. Debug with AI assistance

4. Refactor for understanding

5. Build something original

"AI didn't write my code. It taught me how to write code. Every line in production was understood, tested, and refined by me."

Session Patterns

391

Sessions over 100 messages

1,117

Longest single session

Bimodal

Short check-ins or deep dives

"Bimodal distribution: short check-ins (<15min) and deep sessions (2-6 hours). Middle-length sessions (30-90min) are rare."

Daily Practice Routine

Morning

2 hours exploring new concepts

Afternoon

4 hours building projects

Evening

2 hours refactoring, learning from mistakes

Coded every single day for nearly 3 years. Consistency over intensity.

Skills Mastered

Languages

Python95%
JavaScript95%
TypeScript90%
SQL80%

Core

React, Next.js, Tailwind, FastAPI, Supabase, PostgreSQL, PostGIS

AI/ML

OpenAI, Anthropic, Embeddings, Classification, Fine-tuning

Patterns

Microservices, Event-Driven, SOLID, ETL, Pub/Sub

Code Quality Evolution

Year 1 (Learning)

• Functions per file: 20+ (learning stage)

• Code duplication: 30%

• Documentation: Minimal

• Testing: None

Year 3 (Now)

• Functions per file: 5-8 (modular)

• Code duplication: 15%

• Documentation: Comprehensive

• Testing: Growing (target 70%)

Production Systems Built

Enterprise Tax Credit Platform

$500K+ in credits processed · 10K+ businesses served

YouTube Content Intelligence

32,579 videos processed · Whisper, GPT-4, embeddings

Document Processing Automation

50K+ documents · 7 languages · 98% extraction rate

Real-Time Chat Analysis

11,494 conversations · 36 classifiers · 94% accuracy

Gmail Automation Suite

100K+ emails · 231 specialized processors · 15 providers

What Worked

Daily Consistency — Never missed a day of coding
Project-Based Learning — Built real solutions to real problems
Progressive Complexity — Started simple, added layers
Immediate Application — Used new concepts same day
Error Embracing — Bugs were teachers, not failures

The AI Advantage

"AI accelerated my learning by 10x. Not by writing code for me, but by explaining concepts instantly, debugging alongside me, suggesting better patterns, reviewing code tirelessly, and answering 'stupid' questions without judgment."

Neurodivergent + AI Fluency

"When I speak to you, or any other LLMs, I can communicate at the speed of thought. When I start having to translate everything down for other people, it becomes incredibly hard."

Domain Absorption

"People often think I'm in their field of work. I am able to get close to domain level understanding in a very, very short time."

"I can very, very quickly absorb domain understanding of topics. For as long as I can remember, ideas have just come into my mind like sparks and I've been able to apply first principle thinking to analyze and actualize them."

To Aspiring Developers

"If someone with zero coding experience can write 500K+ lines of production code in under 3 years, you can too."

The key ingredients:

  1. Consistency over intensity
  2. Projects over tutorials
  3. Understanding over copying
  4. Debugging over perfection
  5. Building over planning

The Honest Truth

"Three years ago, I couldn't write 'Hello World'. Today, I architect systems that serve thousands. Tomorrow, I'll build what seems impossible today."

See everything I've built → · How I think → · Full journey documentation →