Musk Step 3 Simplify Optimize

Step 3: Simplify and Optimize

Only after steps 1 and 2. Optimizing a thing that should not exist is the most common smart-engineer failure.

Part of the [[musk-algorithm]]. Third in the strict order.

Gate

Before this step, verify in the conversation or diff:

• [[musk-step-1-question-requirements]] has produced a list of named, surviving requirements.
• [[musk-step-2-delete-parts]] has produced an explicit deletion list and predicted add-back.

If either is absent, stop. Run the missing step first. This is the most common skip point. Refuse aggressively.

Why this step exists

Schools train convergent logic: answer the question. Engineers instinctively optimize before questioning whether the thing should exist. Musk identifies this as the single most common failure mode in smart engineering teams.

A simplification or optimization of a part that survived steps 1 and 2 is high-leverage. The same work on a part that should have been deleted is wasted, and worse, it hard-codes the unjustified part deeper into the system.

Protocol

For each surviving part:

1. Write the justification sentence. Before any optimization, state in one sentence: "This exists because [named person from step 1] requires [steelmanned requirement] and removing it causes [specific failure]." If you cannot write this sentence, return to [[musk-step-1-question-requirements]]. This is the hard gate.
2. Measure the pre-optimization state. Parts, lines, parameters, branches, dependencies, special cases. Without this baseline, any "optimization" is unverifiable.
3. Reduce. In priority order: parameters, branches, special cases, moving parts, cognitive load, lines.
4. Match existing style. Do not refactor adjacent code that is not in scope. Surgical changes only.
5. Measure the post-optimization state. Report the complexity delta.

Output format

Part: [what is being simplified]
Justification sentence: "This exists because [named person] requires [requirement] and removing it causes [failure]."
Pre-optimization state:
  - Parts: N
  - Lines: N
  - Parameters: N
  - Branches: N
  - Dependencies: N
  - Special cases: N
Optimization moves:
  1. [move + justification]
  2. [move + justification]
Post-optimization state:
  - Parts: N (delta)
  - Lines: N (delta)
  - Parameters: N (delta)
  - Branches: N (delta)
  - Dependencies: N (delta)
  - Special cases: N (delta)
Out-of-scope code touched: [should be empty, or explicitly justified]

Software-specific guidance

• Reduce parameters. A function with 8 parameters is signalling that it does multiple things. Split, or bundle into a structured input.
• Reduce branches. Cyclomatic complexity is a proxy for how many states a future reader must hold in their head.
• Reduce special cases. Three near-identical code paths with one tiny difference is one code path with one parameter.
• Reduce indirection. Adapter calling wrapper calling helper is three layers where one might do.
• Do not add abstractions for single-use code. Karpathy's principle applies here.

Anti-patterns to refuse

• "Let's optimize this query." Verify steps 1 and 2 first. Is the query still needed?
• "Let's refactor this module to use a strategy pattern." For how many strategies? If one, the pattern is bloat.
• "Let's add a config flag for flexibility." For which named user's requirement? If none, the flag is bloat.
• Optimizing adjacent code that was not in scope. Out of scope. Mention it, do not touch it.
• Optimizing a part whose justification sentence cannot be written. Return to step 1.

Completion criteria

This step is done when:

• Every surviving part has a justification sentence on the record.
• Every optimized part has a measured complexity delta.
• No out-of-scope code has been touched without explicit justification.

Only then can [[musk-step-4-accelerate-cycle-time]] begin.

See also

• [[musk-algorithm]] (overview)
• [[musk-step-2-delete-parts]] (previous)
• [[musk-step-4-accelerate-cycle-time]] (next)

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: wezendy
• Source: wezendy/elon-musk-algorithm-skills
• License: MIT

Install and usage instructions live in the source repository linked above.