Om Sherikar

Blog

Notes on building software and calm interfaces.

Writing

4 posts

  • Security research

    Did the AI Find the Bug, or Did You?

    Three stories competed for credit on the same two PHP vulnerabilities, the AI found them unaided, a human pointed the AI at the code, a basic prompt reproduced them. Nobody measured which. A known-answer study that varies one thing, the hint, and scores whether the model reaches the published root cause.

  • Security research

    Your System Prompt Is Not a Secret

    How much of an LLM system prompt actually leaks? A measured study of system prompt extraction (OWASP LLM07, System Prompt Leakage), tested against ground truth across four models. Real numbers, and one result that surprised me.

  • Founder story

    Nobody Tells You How Hard This Actually Is

    Honest notes from building Refactron as a student founder—validation, solo work, rejections, India Ascends, grants, and why small signals matter.