The Kryotta blog
Guides and comparisons on working with every AI model — Claude, Gemini, Llama, Mistral, DeepSeek and more — plus what we're building.
I used Auto routing for 220 eBay returns after Boxing Day: which model it picked per query
I processed 220 Boxing Day returns through Auto routing to see which model it picked for each query type—and whether it could handle policy edge cases without manual intervention.
I used 4 prompt templates to write 150 Royal Mail delay emails: which formula per situation
I tested 12 prompt formulas for handling Royal Mail delays and found 4 that work. Here's which template to use for each situation—and why the others failed.
I used 4 prompt templates to reply to 150 Paystack dispute emails: which formula per case
I handled 150 Paystack dispute emails using 4 prompt templates—one for each dispute type. Here's which formula worked for duplicate charges, failed transactions, chargebacks, and delivery complaints.
I used 6 prompt templates for 200 customer emails, invoices and Slack replies: which formula per task
I tested six prompt templates across 200 business tasks—customer emails, invoices, code reviews—and discovered that task type matters more than template elegance. Here's which formula works for what.
I used Claude and Gemini to write 40 M-Pesa refund messages: which one customers actually replied to
I tested Claude and Gemini on 40 real M-Pesa refund messages for my phone accessory shop. One model got more replies and fewer confused follow-ups—here's what each did better.