Its a skill issue!
retained text
My weekly reset was less than 24 hours away and I still had some bandwidth that would go to waste - so I decided to waste my tokens in producing this visual of my usage this past week. Exact prompt used
People on Reddit complain a lot about Claude usage being insufficient - can you create a visual for me that I can share showing how my 20x Max plan which resets Sunday 2PM PT is at 80% total usage and around 50% fable usage (pull the exact numbers from usage stats)
the total number of sessions we have done from last Sunday 2PM
Average duration of each session
Total duration of all sessions
Breakdown of token usage by models
Total cost if this was Charged as Claude API
Some interesting metric around the impact of work done (I don't want to give away what we are building)
Total number of subagents used
Any other interesting metric you find
Create a really nice visual
Feel free to add tips around how we have structured the run sessions and agent setup so people can learn something
I want a kickass, informative and true to data visual - Its okay to generate multiple visuals since 1 long graphic is hard to read on phone.
https://preview.redd.it/af6r690a5kqh1.png?width=2800&format=png&auto=webp&s=2d3175c168f6d2589c28f1108e49f2d86c6eab6a
https://preview.redd.it/ebl5991a5kqh1.png?width=2800&format=png&auto=webp&s=2b203a70db4fc7ee7446a3ca1ee00e69d96cc9b9
https://preview.redd.it/8ld22l0a5kqh1.png?width=2800&format=png&auto=webp&s=d2df4d94ce44a139ca55440ab3aaf6e948244524
Few things you will notice -
\- After the reduced limits kicking in on Monday, I made changes to the runs, reducing the frequency how often they run (reduced the main run from every 3 hours to every 5 hours)
\- Introduced Sonnet for very well defined tasks
\- Reduced Fable usage to be reserved for deeply complex tasks (though, I am now sitting with almost 50% wasted Fable, so I am going to adjust it).
\- All my sessions run on xHigh effort level irrespective of the model being used.
So that's the week. If you read all of that and your takeaway is still "Claude is swindling me," I'd gently suggest the call is coming from inside the chat. Specifically the one chat, the one you opened on Monday, that now contains your entire repo with three abandoned approaches, a pasted stack trace the size of a novel, and your top-tier model being asked to rename a variable while hauling all of it along on every single reply. You are not running out of Max. You are paying full freight to remind the model about Tuesday.
Start fresh sessions, let the repo be the memory, send the cheap model to do the cheap work, and you'll be amazed how far "insufficient" goes. document fetched 2026-09-20 00:35 · canonical https://reddit.com/r/ClaudeCode/comments/1wl0myf/its_a_skill_issue