Plain-English definitions for the terms founders need to know, no computer science degree required.
An AI model trained on vast amounts of text, capable of understanding and generating human-like language. The engine inside Claude, GPT-4, and Gemini. When founders say "AI", they usually mean an LLM under the hood.
The practice of writing inputs that guide an AI toward better outputs. Better prompt = better result. Not magic, it's about being precise, providing context, and defining the format you want back.
An AI system that takes actions autonomously, browsing the web, writing and running code, sending emails, to complete a multi-step goal. Unlike a chatbot, an agent acts on your behalf without a human in the loop for each step.
A technique that gives an LLM access to your own data at query time, your docs, database, knowledge base, so answers are grounded in your content, not just the model's training. The core of most "chat with your data" products.
The maximum amount of text an LLM can "see" in one session, both your input and its output count toward this limit. Larger context windows let the model work with more information at once. Measured in tokens.
The basic unit an LLM processes, roughly ¾ of a word, or 4 characters. Models are priced per token (input + output). Knowing this helps you estimate costs and understand why context windows matter.
When an AI confidently states something factually wrong. Not a bug you can fully eliminate, it's a fundamental property of how LLMs work. Use RAG, verification steps, and human review for anything that matters.
Re-training a pre-built model on your own dataset to specialise its behaviour, style, domain, tasks. Expensive and often unnecessary. Try RAG or few-shot prompting first; fine-tune only when those don't cut it.
Using AI to generate large portions of code through natural language prompts, letting developers focus on architecture and intent rather than syntax. Dramatically speeds up development, when done with proper architecture and code review.
A way of converting text into numbers that capture meaning, so "ship fast" and "launch quickly" end up numerically close. The backbone of semantic search and RAG. Stored in a vector database for fast lookup.
The smallest version of a product that delivers real value to real users and lets you test your core hypothesis. Not a rough prototype, a live product. The goal is to learn fast, not to launch something perfect.
The moment your product genuinely solves a problem for a market large enough to build a business on. You feel it: users come back, tell friends, and pull the product out of you. Everything before PMF is a search. Everything after is execution.
The long-term cost of choosing a quick solution now instead of a better one. Some debt is intentional and fine in an MVP, you're buying speed. Unmanaged debt compounds and eventually kills velocity. The trick is knowing which debt to take on.
A developer who covers both frontend (what users see) and backend (servers, databases, APIs). A full-stack engineer can ship an entire product solo, which is why I hire for it at early stage: one person, full ownership, no handoffs.
A defined contract for how two software systems communicate. When you "connect" two tools, your CRM to your email, your app to Stripe, you're using an API. Think of it as a waiter: takes your order, returns the result.
Software delivered over the internet on a subscription basis, no download, no install. The dominant B2B business model for good reason: recurring revenue, easy to update, and scalable without shipping physical product.
Your strategy for getting the product in front of the right customers. Covers positioning, channels, pricing, and the motion from first user to growth. Engineering can't save a bad GTM. A good GTM can save mediocre engineering.
The rate at which customers stop using your product. High churn is a signal of weak PMF, not a marketing problem. Reducing churn is often more valuable than acquiring new users, retention is the engine of compounding growth.
A senior engineering leader who works with multiple companies at a fraction of the cost of a full-time hire. Provides the technical strategy, architecture decisions, team guidance, and hands-on build capacity of a CTO, without the $200K salary.
The process of making your code live and accessible to real users. Modern deployment is fast, with the right setup, it's a single command or a git push. Good deployment pipelines let you ship confidently and roll back fast if something breaks.