TypeScript Strict Mode: five runtime failures it prevents

TypeScript Strict Mode is not a style preference. It moves production failures into the editor: missing null values, any leakage, unsafe index access, weak function contracts, and JSON that was never validated. A strict project still needs tests, but it stops many crashes before review. External data should start as unknown, be checked at the boundary, and only then become an internal model.

Useful internal references: Word Counter / Read Time Estimator / PageSpeed / Robots and Sitemap Guide / Tailwind CSS 4 Guide.

1. Null and undefined

TypeScript Strict Mode is not a style preference. It moves production failures into the editor: missing null values, any leakage, unsafe index access, weak function contracts, and JSON that was never validated. A strict project still needs tests, but it stops many crashes before review. External data should start as unknown, be checked at the boundary, and only then become an internal model. The practical rule is to decide behavior before runtime: empty state, fallback value, controlled error, or validated model. This keeps reviews focused and prevents hidden assumptions from reaching users.

2. Any leakage

TypeScript Strict Mode is not a style preference. It moves production failures into the editor: missing null values, any leakage, unsafe index access, weak function contracts, and JSON that was never validated. A strict project still needs tests, but it stops many crashes before review. External data should start as unknown, be checked at the boundary, and only then become an internal model. The practical rule is to decide behavior before runtime: empty state, fallback value, controlled error, or validated model. This keeps reviews focused and prevents hidden assumptions from reaching users.

3. Unsafe indexes

TypeScript Strict Mode is not a style preference. It moves production failures into the editor: missing null values, any leakage, unsafe index access, weak function contracts, and JSON that was never validated. A strict project still needs tests, but it stops many crashes before review. External data should start as unknown, be checked at the boundary, and only then become an internal model. The practical rule is to decide behavior before runtime: empty state, fallback value, controlled error, or validated model. This keeps reviews focused and prevents hidden assumptions from reaching users.

4. Weak contracts

TypeScript Strict Mode is not a style preference. It moves production failures into the editor: missing null values, any leakage, unsafe index access, weak function contracts, and JSON that was never validated. A strict project still needs tests, but it stops many crashes before review. External data should start as unknown, be checked at the boundary, and only then become an internal model. The practical rule is to decide behavior before runtime: empty state, fallback value, controlled error, or validated model. This keeps reviews focused and prevents hidden assumptions from reaching users.

5. Unverified JSON

TypeScript Strict Mode is not a style preference. It moves production failures into the editor: missing null values, any leakage, unsafe index access, weak function contracts, and JSON that was never validated. A strict project still needs tests, but it stops many crashes before review. External data should start as unknown, be checked at the boundary, and only then become an internal model. The practical rule is to decide behavior before runtime: empty state, fallback value, controlled error, or validated model. This keeps reviews focused and prevents hidden assumptions from reaching users.

Rollout checklist

• Enable strict in a branch and group errors by risk.
• Start with null handling and external data boundaries.
• Replace implicit any with unknown plus validation.
• Add explicit return types to public functions.
• Keep temporary casts visible until they are removed.

FAQ

Does Strict Mode slow development?

During migration it can, but it reduces production rework by exposing shape and contract errors earlier.

Is optional chaining enough?

No. It prevents one crash but does not decide the user experience when data is missing.

Do schema libraries matter?

Small boundaries can use handwritten guards. Larger products benefit from shared schemas.