Why this exists

Generic LLM code review is forgettable. The notes are correct, the tone is flat, and you close the tab. Reviewers with a stable voice and persistent memory feel different — they feel like mentorship. The bottleneck in most engineers' growth is not reviewer quality. It's reviewer continuity. A senior who sees one PR cannot tell you you're repeating a mistake from three weeks ago. A fictional senior, with memory, can.

What you get back

  • A defined senior engineer character with a backstory, a name, and a specific personality you chose on purpose.
  • A persistent review on every PR, written in their voice, not a default LLM voice.
  • Callbacks to earlier decisions — "you fought me on this exact pattern in the auth refactor; I was right then too."
  • An evolving narrative about your growth as an engineer over weeks and months.
  • The specific patterns they keep flagging — your tells, your blind spots, the thing you reach for when tired.

When to reach for this pattern

Solo work, where there is no one to review you. Small teams without senior engineers, where everyone is roughly the same level and no one is the adult in the room. Learning a new language or paradigm, where you need someone to tell you which idioms are local and which are imported badly from your last stack. And any deliberate practice loop where you want feedback that compounds — where the reviewer in week twelve knows what they told you in week one.