Definition
Chain-of-Thought
Chain-of-thought is a prompting technique that asks a model to reason through a problem step by step before giving its final answer. By generating intermediate reasoning rather than jumping straight to a conclusion, models perform markedly better on multi-step tasks like math, logic, and complex analysis — the explicit steps act as scratch work that improves the result.
Key takeaways
- Chain-of-thought is a prompting technique that asks a model to reason through a problem step by step before giving its final answer. By generating intermediate reasoning rather than jumping straight to a conclusion, models perform markedly better on multi-step tasks like math, logic, and complex analysis — the explicit steps act as scratch work that improves the result.
- Asked for an answer directly, a model commits to its first token immediately and can stumble on problems that require several linked deductions.
- Planoda applies step-by-step reasoning where it improves accuracy on complex work, while what reaches the user is the reviewable proposal — not the raw deliberation — under propose/approve.
Asked for an answer directly, a model commits to its first token immediately and can stumble on problems that require several linked deductions. Chain-of-thought prompting changes the task: by instructing the model to 'think step by step,' you have it generate the intermediate reasoning first. Each step conditions the next, so the model effectively works the problem out on the page before stating a conclusion, which lifts accuracy on arithmetic, logic, and multi-hop questions.
The technique can be triggered with a simple instruction (zero-shot chain-of-thought) or by showing few-shot examples that include reasoning. A growing class of reasoning models is trained to do this internally, producing extended deliberation before the visible answer without being explicitly asked.
Chain-of-thought has real limits. The written reasoning is not a faithful audit of the model's internals — it can reach a correct answer with flawed-looking steps, or a wrong answer with plausible-looking ones. It also costs more tokens and latency, and the visible reasoning may need to be hidden or trusted carefully rather than shown verbatim to end users.
Planoda applies step-by-step reasoning where it improves accuracy on complex work, while what reaches the user is the reviewable proposal — not the raw deliberation — under propose/approve.
Related terms
- Large Language Model (LLM)A large language model is an AI system trained on vast amounts of text to predict and generate language, enabling it to answer questions, summarize, write, and reason over natural language. LLMs power modern AI assistants and agents. They are probabilistic next-token predictors, which makes them remarkably capable but also prone to confident errors.
- Few-Shot PromptingFew-shot prompting is the technique of including a handful of worked examples in a prompt to show a model exactly what you want before asking it to perform the task. The examples demonstrate the input-output pattern, format, and style, steering the model far more reliably than instructions alone — without any retraining of the underlying model.
- Zero-Shot PromptingZero-shot prompting is asking a model to perform a task using instructions alone, with no examples provided. It relies on the broad capabilities the model learned in training to generalize to a task it was never explicitly shown. It is the simplest, fastest way to prompt, and modern models handle many tasks well from a clear instruction by itself.
- InferenceInference is the act of running a trained model to produce output from new input — the moment a model is actually used, as opposed to trained. Every prompt sent to a language model triggers an inference pass. Inference dominates the running cost and latency of AI features, since it happens on every request while training happens once.
- Hallucination (AI)An AI hallucination is output from a language model that is fluent and confident but factually wrong, fabricated, or unsupported by the model's inputs — invented citations, fake quotes, or made-up details. It arises because models predict plausible text rather than retrieve verified facts, making hallucination an inherent risk that must be mitigated, not a bug that's simply fixed.