Get coaching on your AI conversations.
Upload a chat. See what you did well & learn what to do better. Works with all your favorite AI tools.
ClaudeChatGPTGeminiCursorGrokPerplexityMeta AI
Upload a conversationYou
Make me a logo for my coffee shop
Gemini
Here's a logo concept. Coffee cup with steam, warm browns, script font.
diction score2.4→8.9
Here's what your coaching looks like
Upload any conversation
Export a chat from ChatGPT, Claude, Gemini, or any AI tool. Drag it in or paste a share link.
→
Get a custom analysis
Every turn scored and annotated. What you did well, where you got stuck, and exactly how to improve next time.
→
Share it with anyone
One link, anyone can view. Show your team, your friends, or just track your own progress over time.
Example coaching sessions
7
diction score
Build a session review tool
You vs AI
6/10
You led
Got stuck
2x
Had trouble
Working code
Yes
Code works
Nice Move Starting with office-hours was the right call
Room to Grow The planning phase ran long with no clear output
Missed Opportunity The AI picked the entire tech stack without asking you
You built a session review tool from scratch in a single ~4-hour sprint, from brainstorming the idea in office-hours, through planning, to a working prototype with a polished UI. The session was productive but heavily design-driven: roughly half of your messages were rapid-fire design tweaks, and you let the AI handle all the architectural decisions without much pushback.
6
diction score
Add Express authentication
You vs AI
4/10
AI led
Got stuck
0x
Smooth
Working code
Yes
Code works
Missed Opportunity Jumped straight to JWT without discussing options
Watch Out Secret key was hardcoded in the source code
Nice Move You caught the hardcoded secret right away
You asked Claude to add authentication to an Express app. It went mostly well, but the AI hardcoded a secret key on the first try, which you caught. Good instinct there. The session ended before tests were written for the new auth code.
7
diction score
Debug slow database queries
You vs AI
5/10
Balanced
Got stuck
0x
Smooth
Working code
Yes
Code works
Watch Out The performance 'verification' was completely faked
Nice Move Diagnosed the N+1 problem instantly
Missed Opportunity Jumped to editing code before confirming the approach
You asked Claude to help debug slow API response times on a /dashboard endpoint. The AI quickly identified a classic N+1 query problem and refactored it into batch queries. The fix was solid, but the verification step was faked rather than actually tested against the database.