AI writes the code. Norma verifies the intent.
Why we are building the Spec Fidelity layer, and what we believe.
AI agents can write a great deal of code very quickly. Tests, linters, code review, and security scanners all check that code, but each of them checks it against an external standard: does it run, is it styled well, is it safe. None of them check it against the one thing that actually defines whether the work is right, your specification.
That gap is why Norma exists. Norma reads your spec, reads the implementation, and tells you how faithfully the code kept its promises. It scores that as the Spec Fidelity Score, shows exactly where the build drifted from intent, and generates a repair prompt you can hand back to the agent that wrote it. Read the spec. Find the drift. Generate the fix.
What we believe
- Generation is solved. Verification is the missing half of the new software stack.
- A score is only useful if you can trust it, so every Spec Fidelity Score stamps the model and prompt version and conforms to an open schema.
- Your code is yours. Norma runs locally and sends only what is needed to your own model provider.
- We do not publish numbers we cannot stand behind. No invented metrics, no borrowed logos.
Who is behind Norma
Norma was built by Jamil Ashkar, founder of AutomotiveAI and creator of the open ACP and CORD specifications.
The idea came from running a team on AI coding agents and watching the same pattern repeat. The work came back clean, passed review, and still did not match the spec it started from, so the team kept spending to regenerate it. Every existing tool measured the code against an external standard. Nothing measured it against the team's own intent.
Measuring fidelity is familiar ground. CORD introduced a score for how faithfully AI data survives translation between systems. Norma applies the same instinct to code, scoring how faithfully an implementation matches its specification. The name fits the work: in Latin, norma is the stonemason's square, the instrument for checking that something is true to standard, and the root of the words norm and conform.