A foundation model is a large-scale AI model trained on broad, diverse data and designed to be adapted to many downstream tasks. Adaptation happens via fine-tuning, prompting, or API access. The term was coined by Stanford's Center for Research on Foundation Models in 2021 and now describes GPT-4, Claude, Gemini, Llama, Mistral, and similar models that form the base layer of the modern AI stack. The foundation model is to AI applications what AWS is to web applications: shared infrastructure that powers everything built on top.
What distinguishes foundation models:
Scale: hundreds of billions to trillions of parameters. Trained on hundreds of billions to trillions of tokens of data.
General-purpose training: trained on broa...
A B Corporation is either a private certification from B Lab verifying social and environmental performance, or a benefit-corporation legal status that codifies stakeholder primacy. The two are frequently conflated but distinct: the B Lab certification is private and reputational, requiring an 80+ score on the B Impact Assessment plus public transparency and recertification every 3 years. Benefit corporation status (called PBC in Delaware) is a corporate-law election available in roughly 40 US states (including Delaware, California, Texas, Colorado) that modifies director fiduciary duties to require consideration of stakeholder interests alongside shareholders.
B Lab certification (the private certification): B Lab, a nonprofi...
The legal architecture that holds a startup together. This cluster covers entity types and formation (LLC, C-corp, Delaware), governance (board, officers, fiduciary duty), IP protection (trademark, patent, copyright, work-for-hire), employment law (NDAs, non-competes, employment agreements), commercial contracts (MSA, indemnification, arbitration), and privacy/compliance (SOC 2, GDPR, DPAs). 46 entries.
Founders skip this stuff until they can't. The cost of getting it right early is low; the cost of getting it wrong is brutal at diligence or in court.
An AI agent is an LLM-powered system that plans, uses tools, and takes actions over multiple steps to complete tasks autonomously. Tools include APIs, code execution, web browsing, and file operations. Agents go beyond single-prompt question-and-answer to handle complex workflows requiring reasoning, tool use, and iterative correction. "Agentic AI" is the dominant 2025 frontier for AI applications and the next major capability layer beyond chat. Agents are what happens when LLMs stop just answering questions and start doing things.
What distinguishes agents from simpler LLM applications:
Multi-step reasoning: agents break complex tasks into steps and execute each.
Tool use: agents call APIs, run code, browse the web, query database...
Continuous discovery is the practice of conducting weekly customer touchpoints by the product trio (PM, designer, engineer) to inform ongoing product decisions. It was popularized by Teresa Torres in Continuous Discovery Habits (2021) and is structured around mapping desired outcomes to opportunities to solutions to assumption tests, rather than running discovery in concentrated batches separate from delivery. It is the operational evolution of Marty Cagan's product-discovery framing, focused on making customer evidence a weekly rhythm rather than a project.
The core practices Torres specifies: weekly customer touchpoints (at minimum, a 30-minute conversation with a real customer each week, by the whole trio, not just t...
The discipline of building things people want. This cluster covers product management as a function, strategy and discovery, the agile/scrum/kanban operating frameworks, prioritization methods (RICE, Kano), design and UX, testing (alpha, beta, usability), and launch mechanics. 40 entries.
Product is where insight meets execution. The frameworks here are the operating systems most modern product orgs run on.
Agile is an iterative software development philosophy formalized in the 2001 Agile Manifesto, emphasizing working software, customer collaboration, and responsiveness to change. It rejects rigid up-front planning, contract negotiation, and process compliance, implemented through frameworks like Scrum, Kanban, Extreme Programming (XP), and at large-organization scale, SAFe and LeSS. It is the dominant operating philosophy of modern software product teams and the most-misunderstood word in the discipline.
The Agile Manifesto, written in February 2001 by 17 software practitioners at a ski lodge in Snowbird, Utah, states four values: individuals and interactions over processes and tools; working software over comprehensive documentation; ...
The discipline of turning awareness into customers and customers into expansion. This cluster covers the full marketing funnel from awareness to advocacy, every major acquisition channel (paid, organic, content, referral, influencer, affiliate), the metrics that govern each (conversion, CPA, CPC, ROAS), retention and customer success (NRR, churn, NPS, onboarding), brand (identity, voice, positioning, awareness), and the lifecycle marketing patterns that compound. 62 entries.
This is the cluster where the most cross-discipline vocabulary lives. Founders without marketing backgrounds need every term here.
Token economics is the discipline of understanding and modeling the per-token costs and revenue of AI applications. Tokens (sub-word units of text) are the unit of pricing for most LLM APIs and the basis on which AI application unit economics must be modeled. The model includes input tokens (prompt + context + RAG content), output tokens (model response), cache savings, and the per-query economics that determine whether AI applications are profitable. It's the financial layer beneath every AI application.
What a token is:
Token: sub-word unit of text used by LLMs. Roughly 0.75 English words per token, or 4 characters.
Tokenization examples:
The vocabulary of the AI era of startups. This cluster covers the foundational concepts of modern AI (foundation models, LLMs, generative AI), the architecture and operations that power AI applications (Transformer, training data, fine-tuning, prompt engineering, RAG, context window), the economics that determine viability (inference cost, GPU cost, token economics), the strategic moats AI companies build (data flywheel, AI moat, wrapper vs thick wrapper), the safety considerations (alignment, safety), and current-era terms (multimodal, agents, vibe coding). 22 entries.
This cluster is the freshest in the lexicon. If you're building anything AI-adjacent in 2025, every entry here is operational vocabulary.