Prompt Engineering

Standard evaluations measuring code generation quality: from simple function completion (HumanEval) to resolving real GitHub issues (SWE-bench).

Models that navigate imports, call graphs, type systems, and project structure to generate contextually correct changes spanning multiple files.

Agents that generate code, execute it in a sandbox, read error messages, and iteratively fix bugs until all tests pass — closing the generate-test loop.

Translating English descriptions into working functions, classes, and scripts — the core use case driving AI-assisted software development.

Training models to predict missing code given both the prefix and suffix context, powering the inline autocomplete experience in editors like Copilot and Cursor.

Codex, CodeLlama, StarCoder, DeepSeek Coder — models trained on massive code corpora that understand syntax, APIs, libraries, and programming patterns.

Using one LLM to generate, evaluate, and iteratively optimize prompts for another model, automating the prompt engineering process itself.

Decomposing complex tasks into sequential prompt calls where each step's output feeds as context into the next step's input.

Techniques and instructions that force LLM output into machine-parseable formats for reliable downstream integration with software systems.

The model generates multiple reasoning paths, evaluates each branch, and prunes bad directions — systematic search over the space of possible solutions.

A prompting framework where the model alternates between thinking about what to do (Reason), taking actions (tool calls), and processing observations.

Adding "Let's think step by step" or showing worked reasoning dramatically improves accuracy on math, logic, and multi-step problems.

Relying entirely on the model's pre-trained knowledge and instruction tuning by providing only a clear, specific task description.

Including 2-5 input/output examples directly in the prompt so the model infers the desired pattern and applies it to new inputs without any training.

Hidden instructions prepended to every conversation turn that define persona, rules, output format, tool access, and behavioral boundaries.

The practice of crafting structured prompts that reliably guide LLMs to produce accurate, well-formatted, and useful responses.