PostNitro
Day 4 after signup
Nitro Chat now uses about 75% fewer tokens
What we changed under the hood, the new models, and a launch on Tuesday.
Why it works
- Lead with a concrete improvement, then support it with before-and-after figures so readers can quickly understand the update.
- Translate technical savings into practical usage estimates to make the change easier to evaluate.
- Group related feature and model changes under clear section headings, then give each next step its own call to action.
How it is built
- Personalized greeting
- Token savings comparison
- Usage estimates
- Chat feature updates
- Model changes
- Launch invitation and sign-off
Email copy
Hello [First name] 👋 A quick update on the AI chat in PostNitro. Over the past few days we rebuilt how it talks to the AI models, and the same work now uses about 75% fewer tokens. Here's what we measured on the same test conversation, before and after: In short: generating a carousel went from 6,107 tokens to 847. Editing a slide went from about 6,100 to about 1,400. A 12-turn conversation went from 72,361 to about 18,000. Saying "thanks" used to cost around 6,000 tokens; now it costs nothing. HOW MANY CAROUSELS THAT IS Take a 7–8 slide carousel that you generate and then refine with four edits, a colour change and one question. That's about 8,500 tokens on GPT-6 Luna, which works out to: Starter: about 45 carousels a month Creator+: about 140 a month Team+: about 280 a month Generating a carousel on its own uses about 850 tokens, so if you mostly generate and publish, you'll get many more than this. Claude models use about twice as many tokens per turn, and the larger models count at 1.5× or 2×, so on Claude expect between a half and a quarter of these numbers. ALSO NEW IN THE CHAT Switch models mid-conversation from the chat panel header. Use a lighter model for small edits and a stronger one when you need it. A Stop button to cut a reply short. Asking for AI images on an existing carousel now goes straight to the image tool instead of rebuilding your slides. LATEST MODELS We've moved PostNitro onto the latest models from OpenAI and Anthropic, with GPT-6 Luna as the default. If a saved preset used a retired model (such as GPT-4o or the Claude 3.5 and 4 models), it now uses GPT-6 Luna or Claude Haiku 5.5, and you can pick a different one in the preset. Try it now! ONE MORE THING On Tuesday, 13 October 2026 at 12:01 AM PDT (07:01 UTC), we're launching PostNitro for Agents on Product Hunt: let your AI agent design and publish social posts. Follow the PH Launch If anything in the chat feels off after these changes, just reply to this email. It comes straight to me. Founder PostNitro.ai 2803 Philadelphia Pike, STE B 1208 Claymont, Delaware 19703, United States Unsubscribe or Manage Preferences