Knowledge
Knowledge HubConcepts and strategic researchGlossaryDefinitions across six domainsFrameworksStructured strategy methodsTopicsConnected subject areasLibrary
PlaybooksPractical strategy guidesBooksBooks and long-form researchOperator VaultMember resources and toolkits
Explore Discover, the Knowledge Hub, playbooks, and the Operator Vault.
Glossary
Large language model
A large language model (LLM) is a neural language model — typically a Transformer trained at large scale on text (and often code) — that predicts or generates language and can be adapted to tasks such as answering, drafting, classifying, summarizing, and tool-mediated action. In operator language, “LLM” usually names the generative text model behind chat and agent products.
Why it matters
LLMs changed software economics by making unstructured language a programmable interface. For strategy and operations, the important questions are less “do we have ChatGPT?” and more: which tasks tolerate probabilistic generation, what context must be retrieved or verified, and where human decision rights remain mandatory.
LLMs sit inside the broader foundation-model category. Not every foundation model is an LLM, and not every LLM deployment is an agent. An LLM that only completes text in a sandbox differs from an LLM that plans, calls tools, and updates external systems under an agentic operating design.
Example (enterprise): A legal operations team uses an LLM to draft first-pass contract summaries grounded in retrieved clause libraries. The model accelerates drafting; attorneys retain approval rights, and retrieval plus citation checks constrain unsupported claims that a bare prompt would leave unmanaged.
Mechanisms
- Next-token (or related) prediction at scale — pretraining on large corpora yields general language competence transferable via prompts or fine-tuning.
- In-context learning — task instructions and examples in the prompt steer behavior without weight updates.
- Tool and retrieval extension — external search, databases, and APIs compensate for limited parametric knowledge and stale cutoffs.
Boundaries
- LLM ≠ foundation model as exact synonyms; LLMs are a prominent foundation-model subclass focused on language.
- LLM ≠ AI agent. An agent uses models (often LLMs) plus planning, tools, memory, and control loops — and some agents do not use LLMs at all.
- Fluent output is not verified knowledge; enterprise use requires grounding, evaluation, and human checkpoints for consequential decisions.
Related topics
None linked yet.
Related concepts & frameworks
Related Discover collections
Further research
No related resources yet.