How to Bridge the Information Gap with Claude’s Fable 5: A Practical Guide

Understanding the Information Gap

Claude’s Fable 5 is powerful, yet many users experience a mismatch between the prompts, skills, context they provide and the actual task execution. The author calls this mismatch an information gap or “unknowns”.

What Are Unknowns?

The author categorises unknowns into four groups:

Known‑knowns : the explicit instructions written in the prompt.

Known‑unknowns : aspects the user recognises they do not understand.

Unknown‑knowns : details that are obvious to the user once they see them, but are never written down.

Unknown‑unknowns : completely unforeseen factors that may affect the outcome.

Reducing and planning for unknowns is presented as the core skill of “agent‑style programming”.

Helping Claude Help You

Effective prompting requires a balance: overly specific prompts lock Claude into a sub‑optimal path, while overly vague prompts let it fall back on generic best practices that may not fit the task. Providing sufficient context—current step, familiarity with the codebase, and a clear intent—enables Claude to act as a thinking partner.

Pre‑Implementation Strategies

Blind‑spot Scan

Identify potential “unknown‑unknowns” before starting work. Example prompt:

“I’m adding a new authentication provider but know nothing about the existing auth module. Can you perform a blind‑spot pass and list relevant unknown‑unknowns?”

“I’m unfamiliar with colour grading for video. What unknown‑unknowns should I consider?”

Brainstorming & Prototyping

When many “unknown‑knowns” exist, collaborate with Claude to generate ideas and quick prototypes. Example prompt:

“Create an HTML dashboard with four wildly different design directions for me to review.”

“Generate a mock HTML toolbar for a new editor before wiring it up.”

Counter‑questioning

After brainstorming, ask Claude to interview you about ambiguous points, focusing on questions whose answers could change the architecture.

Reference Materials

If you cannot articulate a requirement, point Claude to concrete artifacts such as source code, diagrams, or documentation. Example prompt:

“Read the Rust crate vendor/rate-limiter and re‑implement its retry semantics in our TypeScript client.”

Implementation Planning

Before coding, ask Claude to draft an implementation plan that highlights likely change points (data models, type interfaces, UX flows). This surfaces hidden unknowns early.

In‑Implementation Notes

During execution, maintain a temporary implementation-notes.md (or .html) file to record decisions and deviations. Example prompt:

“Maintain an implementation-notes.md file. If a boundary case forces a deviation, record the reason under ‘Deviations’ and continue.”

Post‑Implementation Activities

Pitch & Explanation Docs

Compile prototypes, specifications, and notes into a concise document (e.g., a Slack‑ready HTML file with a GIF) to accelerate reviewer understanding and approval.

Testing

After a long session, ask Claude to generate an HTML report summarising changes, context, and intuition, and to attach a quiz that you must pass before merging.

Case Study: Publishing a Fable 5 Video

The author used Claude to edit a release video, transcribe with Whisper, prototype a UI synchronised to spoken words using Remotion, and finally asked Claude to teach colour‑grading rather than blindly generate versions, thereby uncovering unknowns about visual quality.

Aligning Map and Domain

Stronger models enable more tasks, but long‑running failures often signal missing unknowns or insufficient planning. Low‑cost activities—documentation, brainstorming, interviews, prototypes, and reference checks—help surface hidden issues before they become expensive fixes.

Before starting any new project, the recommended first step is to let Claude help you discover your own unknowns.