Advanced Prompting for Reasoning Models and Long Context
For experienced prompt and LLM engineers who need to get dependable, cost-controlled results from reasoning models and very long context windows in production systems.
The curriculum is being prepared and reviewed. Enrollment is not open yet.
About this course
Reasoning models and million-token-class context windows reward different habits from the ones prompt engineers learned on earlier models. Step-by-step instructions that once helped can now waste tokens or constrain a model that plans better on its own; dumping an entire corpus into context is possible but not free, and recall across very long inputs is uneven. This course teaches the practices that hold up, grounded in each vendor's published guidance as of October 2026.
The first half covers reasoning models: how hidden or summarized reasoning works, the controls each API exposes (OpenAI reasoning effort, Anthropic adaptive thinking and effort, Gemini thinking levels), how to write goal-and-constraint prompts instead of procedures, how to preserve reasoning across tool calls in agent loops, and how to verify outputs with self-consistency, checkers and graders. The second half covers long context: how attention and position affect recall, placement and query-at-the-end patterns, building your own needle and multi-hop recall tests, prompt caching economics, compaction and memory for long-running agents, and the decision between long context and retrieval. A capstone tunes a long-document analysis pipeline for accuracy, latency and cost. The course ends with an automated exam and a Wirewalk Institute certificate of completion and assessment.
What you will be able to do
Explain how reasoning models allocate test-time computation and how that changes what a prompt should contain
Configure reasoning effort and thinking controls across OpenAI, Anthropic and Gemini APIs and justify the setting against quality, latency and cost
Rewrite procedural prompts as goal-and-constraint prompts and preserve reasoning state correctly across multi-turn tool use
Verify reasoning-model output with self-consistency, programmatic checkers and rubric graders
Design long-context prompts with deliberate placement, document structure and citation requirements, and measure recall with custom needle and multi-hop tests
Apply prompt caching, compaction and external memory to control cost and context growth in long-running sessions
Decide between long context, retrieval and hybrid designs using evaluation evidence rather than window size
Syllabus
6 modules · 20 lessons · about 18 hours
1
Module 1: How Reasoning Models Think
Test-time computation and what it means for the prompt author.3 lessons
Test-Time Compute: What Changes When a Model Reasons Before Answering
Visible, Summarized and Hidden Reasoning: What You Can and Cannot Inspect
Effort and Thinking Controls: OpenAI, Anthropic and Gemini Compared
2
Module 2: Prompting Reasoning Models
Write prompts that direct a planner rather than script it.4 lessons
Goals and Constraints Instead of Step-by-Step Procedures
When Few-Shot Examples and Chain-of-Thought Instructions Hurt
Reasoning Across Tool Calls: Preserving State in Agent Loops
Overthinking and Underthinking: Calibrating Effort per Task
3
Module 3: Verifying Reasoning Output
Trust results because they were checked, not because they look thorough.3 lessons
Self-Consistency, Multiple Samples and Majority Voting
Programmatic Checkers, Unit Tests and Verifier Models
Rubric Graders and Pairwise Comparison for Open-Ended Reasoning
4
Module 4: Long Context Mechanics
Why a bigger window does not mean uniform recall.4 lessons
Position, Attention and the Lost-in-the-Middle Effect
Document Placement, Tagging and Query-at-the-End Patterns
Building Needle, Multi-Hop and Aggregation Recall Tests for Your Own Data
Citations and Quote-First Answering over Long Inputs
5
Module 5: Context Economics and Memory
Keep long sessions fast and affordable as context grows.3 lessons
Prompt Caching: Prefix Design, Cache Hits and Cost Modeling
Compaction, Summarization and Context Editing for Long-Running Agents
External Memory and Scratchpads: What to Persist Outside the Window
6
Module 6: Design Decisions and Capstone
Choose the architecture with evidence and tune a real pipeline end to end.3 lessons
Long Context vs Retrieval vs Hybrid: An Evaluation-Driven Decision
Regression-Testing Prompts Across Model Upgrades
Capstone: Tuning a Long-Document Analysis Pipeline for Accuracy, Latency and Cost
Exam and certificate
The exam unlocks once you have completed every lesson. It draws 25 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 50 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.
Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.
The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by OpenAI, Anthropic, or Google. Product and company names are trademarks of their respective owners.
A free one-hour course built around one realistic case: a resident services coordinator clearing a 412-message backlog with her company's approved AI tool, who nearly mails 63 letters with a legal deadline the assistant got wrong. You learn to write a four-part brief, steer drafts with examples, check what matters and keep restricted data out of the chat box.
Practice task contracts, relevant context, examples, structured output and targeted revision. Compare complete authored outputs against source references and whole-task effort; no paid AI account is required.
Go beyond chat tricks: learn how production prompts are structured, how to decide what goes into the context window, how to get schema-valid JSON, and how to test, version and protect prompts. Every technique comes with before-and-after examples and is checked against the official OpenAI, Anthropic and Google prompting guidance as of October 2026.