High-tech Terms

The language of startups, funding, AI and modern infra - in plain English. Two hundred words, defined simply, so no jargon ever stands between a dreamer and their dream.

Startup, Funding & Investment

Startup: A newly established company designed to grow rapidly and scale.
Venture: A high-potential business project or startup.
Venture Studio: A company that systematically builds and funds multiple startups in-house.
Unicorn: A private startup valued at over $1 billion.
Decacorn: A startup valued at over $10 billion.
MVP (Minimum Viable Product): The simplest version of a product built to test an idea that delivers real value.
Pivot: Changing the core business model or product direction based on feedback.
Burn Rate: The amount of money a company spends per month.
Runway: The number of months a company can operate before running out of cash.
Cap Table: A spreadsheet showing who owns what percentage of the company.
Equity: Ownership stake in a company, usually in the form of shares.
Dilution: When an owner’s percentage stake decreases due to new shares being issued.
Pre-Money Valuation: The value of a company before receiving new investment.
Post-Money Valuation: The value of a company after a new investment round.
Term Sheet: A non-binding document outlining the key terms of an investment.
Due Diligence: The process where investors thoroughly investigate a company before investing.
Exit: When investors realize returns, usually through acquisition or IPO.
Acquisition: One company buys another company.
IPO (Initial Public Offering): When a company lists its shares on a stock exchange.
Bootstrapped: Building and growing a company without external venture funding.
Pre-Seed: Very early funding, usually from founders, friends, and family.
Seed Round: First significant funding round to build and test the product.
Series A: Funding round after product-market fit is demonstrated.
Series B/C: Later-stage funding for scaling the business.
SAFE: Simple Agreement for Future Equity (a popular startup investment contract).
Convertible Note: Short-term debt that converts into equity in a future round.
Angel Investor: A wealthy individual who invests their own money in early-stage startups.
Venture Capital (VC): Professional firms that invest other people’s money in startups.
Syndicate: A group of investors pooling money to invest together.
Lead Investor: The main investor who negotiates terms and leads a funding round.
Pro-Rata Rights: The right for existing investors to maintain their ownership percentage in future rounds.
Liquidation Preference: A priority right for investors to get their money back first in an exit.
Vesting: Gradual earning of equity over time (typically 4 years).
Cliff: A waiting period before any equity starts vesting (usually 1 year).
Bridge Round: Short-term funding to extend runway until the next major round.
Cash Flow: The movement of money into and out of the business.
Unit Economics: Profitability analysis per customer or transaction.
CAC (Customer Acquisition Cost): How much it costs to acquire a new customer.
LTV (Lifetime Value): Total revenue a customer generates over their lifetime.
Churn: The rate at which customers stop using the product.
Valuation Cap: The maximum valuation at which a SAFE or note can convert.
Discount Rate: A discount given to early investors on the next round’s valuation.
Down Round: Raising money at a lower valuation than the previous round.
Up Round: Raising money at a higher valuation than before.
ARR (Annual Recurring Revenue): Predictable yearly revenue from subscriptions.
MRR (Monthly Recurring Revenue): Predictable monthly revenue.
Gross Margin: Revenue minus the direct cost of delivering the product.
EBITDA: Earnings before interest, taxes, depreciation, and amortization.
Board Seat: The right for an investor to join the company’s board of directors.
Information Rights: An investor’s right to receive regular financial updates.
Drag-Along Rights: Allows majority shareholders to force the minority to join a sale.
Anti-Dilution Protection: Protects investors from heavy dilution in down rounds.
Right of First Refusal (ROFR): The right to match any third-party offer to buy shares.
No-Shop Clause: Prevents the company from seeking other investors for a period.
Secondary Sale: Selling existing shares before an IPO or exit.
Tender Offer: An offer to buy shares from existing shareholders.
Strategic Investor: A corporate investor who brings more than just money.
Corporate Venture Capital (CVC): The investment arm of a large corporation.
Accelerator: A program that provides funding, mentorship, and demo days.
Incubator: An organization that helps early startups develop their business.

Growth, Strategy & Advanced Terms

Power Law: Returns in venture capital come mostly from a few big winners.
Portfolio Approach: Investing in many companies expecting most will fail.
Moat: A sustainable competitive advantage that protects a business.
Network Effects: A product becomes more valuable as more people use it.
Flywheel: A self-reinforcing growth loop.
Blitzscaling: Rapid, aggressive growth, even at high cost.
Product-Market Fit (PMF): When a product satisfies strong market demand.
Go-To-Market (GTM): The strategy to sell and deliver a product.
TAM / SAM / SOM: Total, Serviceable, and Obtainable Market sizes.
Traction: Measurable evidence of growth and product-market fit.
Founder-Market Fit: When founders are perfectly suited for their market.
Viral Coefficient: How many new users each existing user brings.
Retention Curve: How many users continue using the product over time.
Cohort Analysis: Analyzing groups of users acquired at the same time.
North Star Metric: The single most important growth metric.
OKRs: Objectives and Key Results (a goal-setting framework).
Equity Compensation: Paying employees with company stock or options.
ESOP: Employee Stock Ownership Plan.
Technical Debt: The cost of choosing quick solutions over better long-term ones.
Growth Hacking: Creative, low-cost ways to grow rapidly.
Vertical SaaS: Software for a specific industry.
Marketplace: A platform connecting buyers and sellers.
Two-Sided Marketplace: A platform serving both buyers and sellers.
Defensibility: How hard it is for competitors to copy the business.
Token Economy: A business model built around crypto tokens.
Web3: A decentralized internet built on blockchain.
DAO: Decentralized Autonomous Organization.
AI-Native: A company built from the ground up around AI.
Synergy: A combined effect greater than the sum of its parts.
Scalability: The ability to grow without a proportional cost increase.
Cash Flow Positive: The company generates more cash than it spends.
Break-even: The point where revenue equals expenses.
SPAC: Special Purpose Acquisition Company (an alternative to an IPO).
Direct Listing: Going public without raising new capital.
Secondary Market: Trading of private company shares.
Lead Time: The time between starting and completing a process.
Gross Burn: Total monthly expenses.
Net Burn: Monthly cash loss after revenue.
Exit Multiple: Valuation expressed as a multiple of revenue or profit at exit.
Liquidity: The ability to quickly convert assets (like equity or shares) into cash without significantly affecting the price.

AI, Modern Tech & Infrastructure

LLM (Large Language Model): A neural network trained on huge text data that generates and reasons over language.
Foundation Model: A general-purpose AI model that many downstream products are built on.
Fine-tuning: Training a base model further on your own data so it specialises for your task.
Prompt Engineering: Crafting the input to an AI model to get a reliably better output.
System Prompt: The hidden instructions an AI model sees before every user message.
Context Window: How much text an LLM can read at once before it forgets earlier parts.
Token: The unit an LLM reads and bills on - roughly 3-4 characters or 0.75 words.
Embedding: A numeric fingerprint of a piece of text that lets a machine compare meaning.
Vector Database: A store of embeddings used to find similar items by meaning, not keywords.
RAG (Retrieval-Augmented Generation): Letting an LLM look things up before it answers so it stays grounded.
Hallucination: When an AI confidently invents a fact that is not true.
Grounding: Forcing an AI to cite or stay within a known source so it cannot invent.
Agent: An AI that picks its own tools and runs multi-step actions to reach a goal.
Tool Use: When an LLM calls real software functions instead of just writing text.
Function Calling: A protocol that lets an LLM output a structured request for a tool to run.
MCP (Model Context Protocol): A standard for connecting AI agents to external tools, data, and apps.
Chain-of-Thought: A prompt technique that asks the model to reason step by step out loud.
Reasoning Model: An LLM trained to think for longer before answering, like o1 or Claude 3.7.
Multimodal: An AI that handles more than text - images, audio, video, code together.
Diffusion Model: The technique behind most modern image generators - turns noise into pictures.
Transformer: The neural network architecture behind nearly every modern LLM.
Inference: Running a trained model to produce an answer (as opposed to training it).
Training Run: The expensive process of teaching a model from scratch on huge data.
Distillation: Compressing a big model into a smaller, cheaper one that behaves almost the same.
Quantisation: Shrinking a model by storing its weights in lower precision so it runs faster.
Alignment: Making an AI behave the way humans actually want it to.
Guardrails: Filters and rules around an AI to keep it from doing harmful things.
Red Teaming: Adversarially probing an AI to find ways it breaks before users do.
Eval: A repeatable test that measures whether an AI is getting better or worse.
AI Gateway: A single endpoint that routes requests to many model providers with fallbacks.
OpenAI-compatible Endpoint: A non-OpenAI API that speaks the same protocol so existing tools just work.
BYOK (Bring Your Own Key): A product pattern where the user supplies their own API key per call.
Cost-per-1k-tokens: The standard pricing unit for LLM usage.
Prompt Caching: Reusing the cached result of a long system prompt to save tokens and time.
Streaming: Sending an AI response token-by-token so the user sees it as it generates.
Fine-grained Permissions: Letting an AI agent take only the exact actions you have approved.
Sandbox: An isolated environment where untrusted code or AI actions can run safely.
Vibe Coding: Building software primarily by describing intent to an AI rather than typing code.
Spec-driven Development: Letting an AI agent write code from a written specification, then verifying.
Generative UI: Interfaces that an AI assembles on the fly based on the task at hand.
Synthetic Data: AI-generated training examples used when real data is scarce or private.
Knowledge Graph: A structured map of entities and relationships, often used to ground AI.
Edge Inference: Running an AI model directly on the user device instead of in the cloud.
Latency: How long the user waits between asking and seeing an answer.
TTFT (Time to First Token): How long an LLM takes before its very first character appears.
Throughput: How many tokens or requests a model can serve per second.
Cold Start: The slow first hit on a serverless function before its instance is warm.
Serverless: Running code without managing servers - you pay per invocation.
Fluid Compute: Vercel's model where one function instance handles many concurrent requests.
Edge Function: A small piece of code run on a CDN node close to the user.
CDN (Content Delivery Network): A global network of servers that caches files close to the user.
Webhook: A URL another service hits when something happens, so your app can react.
Polling: Repeatedly asking an API for updates instead of being notified.
SSE (Server-Sent Events): A one-way stream that pushes updates from the server to the browser.
WebSocket: A two-way live connection between browser and server.
Idempotent: An operation that produces the same result no matter how many times you run it.
Race Condition: A bug where two processes interfere depending on which finishes first.
Optimistic UI: Showing the user a result before the server confirms it, for snappier apps.
Hydration: When a static HTML page becomes interactive after JavaScript loads.
SSR (Server-Side Rendering): Generating each page on the server per request for fresh content.
SSG (Static Site Generation): Building all pages once at deploy time so they load instantly.
ISR (Incremental Static Regeneration): Mostly-static pages that quietly rebuild themselves on a schedule.
PPR (Partial Prerendering): A page where the shell is static and only the dynamic parts stream in.
API Route: A backend endpoint that lives inside the same project as the frontend.
Route Handler: A Next.js file that responds to a single API URL.
Server Action: A function the browser can call directly without writing a separate endpoint.
JWT (JSON Web Token): A signed token used to prove who you are between browser and server.
OAuth: A standard that lets users grant apps access without sharing passwords.
SAML: An enterprise sign-on protocol used for company-wide single sign-on.
OIDC (OpenID Connect): A modern identity layer built on top of OAuth.
SSO (Single Sign-On): One login that unlocks several apps.
2FA / MFA: Requiring a second proof beyond a password (code, key, biometric).
Passkey: A passwordless login backed by your device, replacing typed passwords.
Rate Limit: A cap on how often a client can hit an API in a given window.
Backoff: Waiting progressively longer between retries so you do not hammer a failing service.
Circuit Breaker: A pattern that stops calling a failing service until it recovers.
Observability: Being able to see why a live system did what it did via logs, traces and metrics.
Telemetry: The measurements an app emits about its own health and behaviour.
Feature Flag: A switch that turns part of the product on or off without redeploying.
A/B Test: Comparing two versions of something to see which one users prefer.
Cohort: A group of users who joined or behaved the same way, tracked over time.
Retention Curve: How many users keep using the product over days, weeks, months.
PMF (Product-Market Fit): When users genuinely pull the product out of your hands.
GTM (Go-to-Market): The plan for how a venture reaches its first customers.
ICP (Ideal Customer Profile): A precise description of the customer the product is built for.
Wedge: The narrow, sharp use case a product uses to break into a market.
Moat: Whatever durable advantage stops a competitor from copying you.
TAM / SAM / SOM: Total, serveable, and obtainable market sizes for an opportunity.
NPS (Net Promoter Score): A standard survey question that scores how likely users are to recommend you.
DAU / MAU: Daily and monthly active users - the two most-cited usage metrics.
Activation: The moment a new user does the thing that proves they got value.
North-Star Metric: The one number every team agrees the product should move.
Web3: Apps that run on public blockchains where users own their data and keys.
Smart Contract: Code that lives on a blockchain and runs the same for everyone, every time.
Wallet: The user's container of crypto keys, used to sign transactions.
Stablecoin: A crypto token pegged to a real currency, used for low-volatility payments.
Gas: The fee paid to run a transaction on a blockchain.
On-chain / Off-chain: Data stored on the blockchain itself vs. anywhere else.
DAO (Decentralised Autonomous Organisation): A group whose decisions run on a smart contract instead of a board.
Tokenomics: The economic design of a crypto token - supply, demand, incentives.
Zero-Knowledge Proof: A cryptographic trick that proves something is true without revealing it.
Open Source: Source code published publicly so anyone can read, fork, or contribute.