Urgent.News

What's breaking now, across thousands of outlets.

AI

Freeze a Reader Contract Before Models Draft Tutorial Pages

Freeze a Reader Contract Before Models Draft Tutorial Pages Generated tutorial pages stay trustworthy only when a human freezes audience, prerequisites, and success checks before drafting begins. Models may draft connective prose around those owned facts, but they should not invent the facts themselves. This article defines an ownership split, a contract file, and a checker that rejects drafts…

Freeze a Reader Contract Before Models Draft Tutorial Pages Generated to ensure trustworthy tutorial pages. Models can draft prose, but they shouldn't invent facts. This article defines an ownership split, a contract file, and a checker to reject drafts with unlisted assumptions. Tutorial failures often start with silent assumptions, not awkward wording or missing commas.

A draft may include unverified details like operating system, language version, time budget, or success string, which can mislead newcomers. Diátaxis treats tutorials as learning paths that must remain reliable, so preconditions belong to an owner who can verify them. Authors can refer to the Diátaxis tutorial section for guidance.

The workflow separates reader facts from draftable explanation, ensuring reviewers focus on facts that could mislead newcomers. The method is narrow, leaving command inventories and escalation pages to their existing owners. The table controls the workflow, with humans editing the contract file and models quoting contract lines or drafting marked sections.

A model cannot add prerequisites, durations, versions, platforms, or success strings. If a draft contains new commands, escalations, or limits, it fails closed for human review. Reviewers should reject mixed pages instead of hoping for later edits. Block owners have permission to quote contract lines, allowed items, allowed strings, and disallowed additions.

Step-order rationale involves model drafting without new facts, with humans quoting allowed items, and models drafting if no new claims appear. The process starts by extracting reader facts from trusted files, copying only verifiable facts and recording their sources. Unknown fields remain empty for human review. The contract file includes audience, goal, sources, allowed platforms, allowed prerequisites, forbidden additions, success checks, and non-goals.

The tutorial skeleton matches draftable and quote-only lists from the contract, with contract lines under quote-only headings and draftable headings empty or with placeholders. The model receives the contract, skeleton, and limits against new facts. If the model returns new headings or prerequisites, the output is discarded, and the process is rerun with the same frozen skeleton.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

It's Not the Model, It's You

Every time a new frontier model comes out, we go through the same rituals: benchmarks get reported, leaderboards get reshuffled, and the social feeds fill up with a heady cocktail of vibes, hot takes…

More from Sunday 11 October →