Index / Writing
Notes on building things.
Business, startups, product, software architecture and applied AI. 105 posts. Subscribe via RSS.
Series
13Startup Basics · 14 partsStart: Default Alive or Default Dead? Runway Math for Small Startups
- Default Alive or Default Dead? Runway Math for Small Startups
- Your First 10 Customers Won't Come From Marketing
- Unit Economics in Plain English: CAC, LTV, Payback and Margin
- Validate an Idea in One Week
- Product-Market Fit: How to Know You Have It
- Cohort Retention: The Chart That Tells the Truth
- How to Price a SaaS or API: Value Metrics, Tiers and Trials
- Positioning: Who It's For and Why It's Better
- Distribution Beats Product
- Activation: The First Five Minutes
- Founder-Led Sales
- Pivot or Persevere
- Cap Tables in Plain English: Who Owns What, Round by Round
- Your First Hire: When, Who, and What It Really Costs
Hosting, by the Numbers · 5 partsStart: A 10-Unit Host's Morning, Before and After Automation
AI Engineering · 5 partsStart: Build an MCP Server: Give Your AI Agent New Tools
RIAI Lab Notes · 6 partsStart: RIAI: Toward a Terminal Agent That Writes Its Own Skills
- RIAI: Toward a Terminal Agent That Writes Its Own Skills
- A Terminal That Learns What Commands Do
- Writing Its Own Code: From a Teacher Model to Code Built by Search
- Next: Teaching an Agent to Imagine Code With JEPA
- Watching JEPA Write Code Live: What It Got Right, and the Check It Fooled
- What the Research Says: Mapping RIAI Against the Field
How AI Works · 5 partsStart: How ChatGPT Actually Works, Explained Simply
Business Essays · 6 partsStart: People Don't Pay for Features. They Pay to Do Less.
- People Don't Pay for Features. They Pay to Do Less.
- Your Real Competitor Is a Spreadsheet
- Boring Businesses Make the Best Businesses
- AI Won't Replace Small Business Owners. It Will Replace Their Busywork.
- Investors or Independence? When a Small Startup Should Raise
- Why Most Small Service Businesses Charge Too Little
Applied AI · 12 partsStart: The Rise of Generative AI: Simplifying Learning
- The Rise of Generative AI: Simplifying Learning
- Generative AI Large Language Model – Train From Pre-trained Model
- Recommendation System: How Important Is It?
- How to Tell If an AI Product Is Real or Just a Demo
- The Hidden Costs of AI Tools Nobody Shows in the Demo
- What an "AI Agent" Actually Is, Without the Hype
- Why Santara Sends One Briefing, Not a Chatbot
- Stop Getting Generic Answers From AI: The 5-Part Prompt Recipe
- 5 Famous AI Demos That Weren't What They Looked Like
- Using AI With Customer Data: The Traffic Light Rule
- "Built With AI in a Weekend": What Happened Next
- A Chatbot Sold a Car for $1: 5 AI Fails Every Business Should Know
After Vibe Coding · 9 partsStart: Vibe Coding Is Real. So Is the Bill.
The One Sheet · 4 partsStart: The One Google Sheet That Runs a Small Business
The Road to AGI · 8 partsStart: What Does "AGI" Even Mean?
- What Does "AGI" Even Mean?
- Bet #1: Write Down the Rules (Symbolic AI)
- Bet #2: Let It Learn (Neural Networks)
- Bet #3: Reward Is Enough (Reinforcement Learning)
- Bet #4: Just Make It Bigger (Scale)
- Bet #5: Teach It to Think (Reasoning and Agents)
- The Other Bets: Neurosymbolic, World Models, Brains and Evolution
- AGI: Where We Are Now (October 2026)
How AI Works: Advanced · 12 partsStart: Interpretability: How Scientists Look Inside an AI's Mind
- Interpretability: How Scientists Look Inside an AI's Mind
- Mixture of Experts: Why Giant Models Only Use Part of Their Brain
- Long Context and the KV Cache: How AI Reads a Whole Book at Once
- Shrinking AI: Quantization and Distillation Explained
- Why AI Runs on GPUs
- How AI Sees: Vision Transformers and CLIP
- How Image Generators Work: Diffusion Explained
- Glitch Tokens: The Strange Words That Broke GPT-3
- Reward Hacking: When AI Wins the Wrong Game
- AI Benchmarks and Goodhart's Law: Why Leaderboards Mislead
- Synthetic Data and Model Collapse
- World Models: Can AI Simulate Reality?
Founder Lessons · 6 partsStart: My First Business Teacher: William Christopher Skelley
- My First Business Teacher: William Christopher Skelley
- Read Every Word You Sign: Term Sheets for First-Time Founders
- What Steve Jobs Teaches Founders, the Light and the Dark
- Nothing Was Wrong. That Was the Problem.
- Choosing a Co-Founder: What I Checked Before Saying Yes
- Strangers Trading With Strangers: What Chợ Gem Taught Me About Trust
Software Design · 8 partsStart: Clean Architecture in Practice: How to Structure Large Backend Systems
- Clean Architecture in Practice: How to Structure Large Backend Systems
- Writing Better Code with SOLID Principles
- Background Jobs Done Right: Queues, Retries and the Outbox Pattern
- Multi-Tenant SaaS Database Design: Tenant IDs, Isolation and Row-Level Security
- Start With a Modular Monolith, Not Microservices
- Time, Money and IDs: The Data Types That Bite Backend Developers
- Treat the LLM Like an Unreliable API: Designing AI Pipelines That Hold Up
- Integrating Airbnb, Booking.com and WhatsApp: Lessons From Messy APIs
Latest
1–12 of 105Your First Hire: When, Who, and What It Really Costs
When a small startup should make its first hire, what it really costs beyond salary, and how to choose with paid work samples instead of puzzles.

Cap Tables in Plain English: Who Owns What, Round by Round
Founder shares, vesting, the option pool, SAFEs, dilution and liquidation preference, with one fictional startup traced from founding to exit.

Pivot or Persevere
How to decide whether to keep going or change direction: Eric Ries's pivot meeting, ten pivot types, kill metrics set in advance, and a worked example.

Founder-Led Sales
Why founders must sell first, a six-step sales process for developers, discovery questions that don't lead, pipeline math, and when to hire.

Activation: The First Five Minutes
Most signups leave before they ever see what your product does. What activation is, how to find your own aha moment in data, and why fixing it beats buying traffic.

Distribution Beats Product
A good product with no way to reach people dies quietly. The 19 traction channels, the Bullseye method, channel math and cheap tests, explained simply.

Positioning: Who It's For and Why It's Better
What positioning is, April Dunford's five components in order, why 'for everyone' fails, how to test it, and a worked example with a before/after headline.

How to Price a SaaS or API: Value Metrics, Tiers and Trials
Choose what you charge per, set a price between cost and value, build good-better-best tiers, pick a trial, and test prices with real customers.

Cohort Retention: The Chart That Tells the Truth
Total-user charts hide churn. How to build and read a cohort retention table, user and revenue retention, with pandas code and Google Sheets formulas.

Product-Market Fit: How to Know You Have It
What product-market fit means, five ways to measure it, the false signals that fool founders, and a worked survey example with segments.

Validate an Idea in One Week
Before you build for months, spend seven days testing the one thing that must be true. A day-by-day plan, pass bars, and how to decide go, pivot or stop.

Why Do Neural Networks Generalize? Overfitting, Double Descent and Grokking
Big networks memorise their data and still work on new data. Overfitting, double descent and grokking in plain English, with a numpy demo you can run.
