Question 1

Why did my startup get a low score?

Accepted Answer

Low scores usually reflect weak market fit, crowded competition, unclear differentiation, or high execution risk. Our AI analyzes your idea across these dimensions and explains the main issues in the breakdown. Use the verdict and suggested pivot to refine your concept and re-roast to see improvement.

Question 2

What do the roast score tiers mean?

Accepted Answer

Ship It (85–100): Strong idea, worth validating further. Decent (70–84): Promising but has gaps to address. Needs Work (50–69): Significant issues; use feedback to iterate. Roasted (0–49): Major concerns; consider pivoting or solving the core problem before building.

Question 3

How does DontBuildThis score startup ideas?

Accepted Answer

Our AI evaluates your idea on market fit, competition, differentiation, execution risk, and clarity. It returns a score out of 100, a tier, a verdict, and a detailed breakdown so you know exactly what to improve before spending time and money building.

Question 4

Can I improve my startup idea score?

Accepted Answer

Yes. Use the breakdown and verdict to address weak points, then submit an updated idea or pivot. Many founders improve their score by narrowing the problem, clarifying the target customer, or strengthening their unique angle.

Question 5

What is a good score for a startup idea?

Accepted Answer

Scores of 70+ indicate promising ideas worth exploring. 85+ suggests strong potential. Scores below 50 mean the idea likely needs significant rework or a pivot before it’s worth building.

Question 6

What score did "Problem: Every knowledge worker drowns in context scattered across Slack, email, GitHub, meeting transcripts, and AI conversations. They try "second brain" tools like Obsidian or Notion, but those require manual input — so everyone gives up after two weeks. Meanwhile, coding agents are getting faster but still make mistakes because they lack context about what you discussed, decided, and know.

Solution: A personal context engine that automatically ingests signals from everywhere you work — Slack, email, git, CI, meeting transcripts, even Claude Code conversations — and uses an LLM to classify, group, and link them into "situations" (things that need your attention). It stores everything as plain markdown in an Obsidian-compatible vault. The context is served wherever you need it: a TUI in your terminal, a browser extension, a mobile app, and — critically — as an MCP server that coding agents can query directly. Your agent asks "what do I need to know about the inventory sync?" and the engine answers with the relevant Slack thread, the CI failure, and last Tuesday's meeting discussion. You never file a ticket, write a note, or brief an agent. The context just exists.

Target market: Developers and knowledge workers using AI coding agents (Claude Code, Cursor, Copilot) who spend more energy gathering and repeating context than making decisions.

Differentiation: Every "second brain" asks you to put things in. We automate ingestion. Every developer dashboard shows you data. We reduce it to decisions. And the MCP angle is unique — we're not just a tool for humans, we're a context layer for AI agents. The agents become our distribution channel." get?

Accepted Answer

"Problem: Every knowledge worker drowns in context scattered across Slack, email, GitHub, meeting transcripts, and AI conversations. They try "second brain" tools like Obsidian or Notion, but those require manual input — so everyone gives up after two weeks. Meanwhile, coding agents are getting faster but still make mistakes because they lack context about what you discussed, decided, and know.

Solution: A personal context engine that automatically ingests signals from everywhere you work — Slack, email, git, CI, meeting transcripts, even Claude Code conversations — and uses an LLM to classify, group, and link them into "situations" (things that need your attention). It stores everything as plain markdown in an Obsidian-compatible vault. The context is served wherever you need it: a TUI in your terminal, a browser extension, a mobile app, and — critically — as an MCP server that coding agents can query directly. Your agent asks "what do I need to know about the inventory sync?" and the engine answers with the relevant Slack thread, the CI failure, and last Tuesday's meeting discussion. You never file a ticket, write a note, or brief an agent. The context just exists.

Target market: Developers and knowledge workers using AI coding agents (Claude Code, Cursor, Copilot) who spend more energy gathering and repeating context than making decisions.

Differentiation: Every "second brain" asks you to put things in. We automate ingestion. Every developer dashboard shows you data. We reduce it to decisions. And the MCP angle is unique — we're not just a tool for humans, we're a context layer for AI agents. The agents become our distribution channel." scored 91/100 (🔥 Ship It). This is the context engine the AI wave actually needs: if you don't build it, someone else will.