KillMyIdea

What it is

KillMyIdea is an open-source proof of concept designed to stress-test TypeSafe's non-generative Jev model. Instead of relying on a traditional conversational LLM that outputs conversational reasoning, it takes an idea pitch and runs roughly 10 targeted evaluation questions in parallel. In a fraction of a second, the system returns a composite numerical score from 0 to 100 along with granular diagnostic classifications (such as "SHIP IT" or "FIX IT").

Core value proposition

• Zero generative chatter: Eliminates conversational fluff and prompt drift by using a deterministic model that outputs solely quantitative probabilities rather than text.
• Sub-second parallel evaluation: Runs all evaluation queries concurrently, delivering comprehensive multi-axis feedback in roughly 500 milliseconds instead of 30 to 60 seconds.
• Weighted multi-factor rubrics: Scores independent axes like Demand, Customer Clarity, and Buildability on fixed 0–4 scales, applying heavier weights to critical factors such as market need and real problems.
• Goal-aware calibration: Supports project goals (such as Open Source or Just for Fun) to prevent penalizing ideas that intentionally avoid monetization.

What it can replace

Current realities and limitations

• Evaluates the pitch, not the market: The underlying model evaluates the descriptive phrasing of the pitch against calibrated linguistic rubrics; it does not query live market data or verify external reality.
• Blind to domain regulations: Specific barriers like regional compliance, licensing, or complex legal hurdles are not automatically factored in unless explicitly detailed in the prompt.
• Lack of qualitative narrative: Because the model returns only numbers and discrete levels, users receive no direct written explanations for why a specific metric received a low score.
• Heuristic weighting: The composite score is derived from an author-defined weighted average rather than an established venture capital benchmark.

Verdict

Primary recommendation Use KillMyIdea as a lightning-fast, objective sanity check to spot weak phrasing or poorly defined target audiences before drafting comprehensive proposals.

KillMyIdea proves that non-generative evaluation models like Jev can slash latency and eliminate non-deterministic variance in automated workflows. While it should not be treated as an authoritative market verdict, it serves as an effective first-pass triage filter for developers, founders, and agent pipelines.

Development lifecycle and iteration loop

The project was implemented as an open-source web application (killmyidea source code) to experiment with direct access to TypeSafe's Jev model. Iterations on community feedback included adding a "Refine Idea" workflow and optional goal parameters to adjust scoring logic for non-commercial projects.

Metrics and monetization

• Pricing: Completely free to use.
• Licensing: Open source under an accessible public repository.
• Performance: Capable of evaluating 10 distinct rubrics concurrently within roughly 539 milliseconds.
• Monetization: No direct monetization; built as a technical demonstration and POC for developer feedback.

Community reception and key debates

• Deterministic predictability: Developers praised Jev's parallel speed and non-generative nature, noting its potential as a deterministic guardrail for agentic architectures.
• Scoring skepticism: Commenters initially questioned whether scores differed meaningfully from pseudo-random numbers, leading the author to clarify the calibrated 0–4 rubric mechanics.
• Data harvesting concerns: Some users hesitated to submit proprietary startup ideas, prompting the author to highlight an explicit opt-out toggle that prevents pitch storage.