The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
Two writing responses can share a total score and require different feedback.
One may fulfil the task with limited grammar. Another may use sophisticated language while missing a required recommendation. A single total hides that difference.
I review the criterion profile first: task achievement, grammar, vocabulary, organisation and register. Feedback should explain the most consequential gap using evidence from the response.
In AI Score Benchmark Pro, a total out of 25 is not automatically converted to a CEFR level. The qualitative level and criterion ratings remain separate judgments.
AI Score Benchmark Pro uses AI-authored provisional synthetic references. Independent human rating has not been completed.
Free preview: https://contra.com/products/UDTrn11c-ai-score-benchmark-pro-free-writing-qa-preview
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
I realized that prompt is just 20-30% mostly, it will require:
Claude Skill
Understanding of brand or feature
Taste
visualization
tell what exactly what you think
Going back and forth with multiple iterations
AI in loop
research paper
articles
rules
collaboration b/t apps
Connectors
Knowledge in sound
Ease graph
Locked reference frames
API
MONEY
etc etc
In the end I want to say I'm still learning and I will share what I made and what I change from the last version
good one
Exploring two hero sections for Timmly 🛠️
A - split with the app
B - centered, just the words
Which one would you ship? 👀
Designed with Claude
6 voted
100%
0 voted
0%
6 votes
Closed
Ship A, Grima. The split with the app gives instant context: visitors see the product in action before they read a word. The centered version looks clean but the app screenshot sells it.
When I scope clinical AI, I separate drafting from rule-based checks.
In Arogya, typed or dictated rounds become editable AI drafts with the original source preserved. Deterministic rules handle checks such as allergy conflicts and abnormal-lab flags. Clinicians review and approve AI content before it reaches the chart.
That separation matters when a model or prompt changes. I'd want the team to rerun the rule checks with the same inputs and inspect their results independently of the draft's wording.
For a client evaluating an AI scribe, I'd ask to see why a flag appears and how the clinician reviews it before approving the record.
Public case: https://kapilchauhan.netlify.app/case/arogya.html
kapilchauhan.netlify.app
Arogya / Aurora EHR: AI for inpatient care | Kapil Chauhan
Production hospital and EHR platform that accelerates rounds, documentation, records, and care operations with clinician-reviewed AI.