← Reddit

I made a ruleset to turn ChatGPT, Claude, Gemini into a CV writer that interviews you

Reddit · Anbeeld · May 12, 2026
A developer created a ruleset for AI chatbots that transforms their approach to CV writing through structured interviews instead of requesting entire career histories at once. The ruleset enforces evidence-based drafting, converts vague accomplishments into specific quantified outcomes, audits for generic language and weasel words, and adapts to existing LinkedIn profiles or starts from scratch. The free tool reportedly produces stronger CVs with significantly less writing burden than traditional methods.

Detailed Analysis

A developer has released an open-source prompt ruleset called RESUME.md on GitHub, designed to transform general-purpose AI assistants — including Claude, ChatGPT, and Gemini — into structured, interview-style CV writers. Rather than relying on a single large prompt dump, the system guides users through their professional history one question at a time, building a complete picture before generating any draft output. The tool accommodates users at different starting points: those with nothing prepared are walked through from scratch, while those who already have a CV or LinkedIn profile can paste it in, allowing the system to lock confirmed facts and probe only for gaps.

Several deliberate design choices distinguish the ruleset from simply asking an AI to "write my CV." The system enforces an evidence-first discipline — it refuses to draft bullet points until it has concrete outcomes or, at minimum, qualitative anchors, explicitly walking users down a hierarchy of specificity when exact metrics are unavailable. A self-auditing pass is built into every draft cycle, automatically flagging weasel verbs, hollow phrases like "dynamic team player," and what the author terms "leaked process language" — the tendency of AI to describe job duties rather than professional impact. Regional market conventions are also embedded, removing the guesswork around norms like profile photos or personal details that vary significantly across hiring markets.

The project speaks directly to a well-documented failure mode of generalist AI use in professional writing: without structured constraints, large language models default to statistically average outputs, producing documents that sound polished but are indistinguishable from thousands of other AI-assisted applications. By encoding domain-specific rules as persistent instructions rather than relying on one-shot prompting, the author effectively creates a lightweight agent layer on top of commodity models. This approach — using detailed rulesets to specialize general-purpose AI for a narrow task — is increasingly common among technically sophisticated users who want professional-grade outputs without paying for purpose-built SaaS tools.

The decision to release the project freely on GitHub, with the explicit acknowledgment that paid services are inaccessible to unemployed job seekers, reflects a broader democratization dynamic in AI tooling. As the capability gap between frontier models narrows, competitive differentiation increasingly shifts to prompt engineering, workflow design, and domain knowledge encoding — skills that are distributable at zero marginal cost through open repositories. RESUME.md is a practical example of this shift: its value lies not in access to any particular model, but in the structured methodology layered on top of models that any user can already access for free. The fact that it works across Claude, ChatGPT, and Gemini simultaneously underscores how model-agnostic well-constructed prompt architectures can be when the underlying task is sufficiently well-defined.

Read original article →