Use grading criteria to improve subjective outputs like design
Define concrete grading criteria based on principles to improve subjective outputs like design quality.
Turn subjective judgments into concrete, gradable terms using a set of criteria.
For subjective tasks like design, establish specific grading criteria that encode principles and preferences. This provides the agent with concrete goals, moving beyond vague notions of 'good design' to measurable attributes.
Instead of asking 'Is this design beautiful?', use criteria like 'Does this follow our principles for good design?' or 'Is the layout original and visually appealing?'
Improve subjective outputs, such as design generation, by defining specific grading criteria based on underlying principles. This transforms vague quality assessments into concrete, measurable attributes that agents can effectively work towards.
It gives agents something concrete to grade against, moving beyond subjective taste to actionable principles. This allows for more consistent and reliable evaluation and iteration.
- Harness design for long-running application development ↗Anthropic Engineering · Anthropic Engineering · Mar 24, 2026 · paraphrased
First seen on Subvolts Mar 24, 2026. Updated Sep 11, 2026. Spotted a problem? Tell us.
- Harness design for long-running application developmentAnthropic · Anthropic Engineering · Mar 24, 2026
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- HeuristicSeparate agent work from agent evaluation for better resultsUse distinct agents for task execution and output evaluation to improve performance and overcome self-evaluation bias. · 10 min
- Anti-patternRelying on a single agent for both generation and self-evaluationAvoid using a single agent for both generating content and evaluating it due to inherent self-praise bias. · 5 min
- HeuristicAI agents are like teammates, not just chatbotsManage AI agents like teammates, maintaining control through visible work and check-ins, rather than just prompting. · 2 min
- Skill or agent configJoplin agent for automated note filing and retrievalAutomate note organization and retrieval in Joplin using a custom agent and Python scripts. · 60 min
- Setting tipStore AI agent traces and evaluation data with projectKeep traces, evaluation datasets, and results with your project for repeatable testing. · 5 min
- WorkflowBuild and test an AI agent using Microsoft Foundry ToolkitCreate, configure, test, and evaluate an AI agent within VS Code. · 60 min
Questions people ask
- Which AI tools does this work with?
- Any current model. It was seen with in the sources.
- Where did this come from?
- Anthropic Engineering (Anthropic Engineering). Subvolts collected, normalized and credited it; the link above opens the original.