JeffersonNunn.eth (@mindragon) keeps 16 Command Code windows going every day. After three weeks, he posted his usage dashboard and a list of what he shipped, and noted he still had plenty of spend left.
What the dashboard shows
JeffersonNunn.eth didn't share a benchmark. He shared his Command Code usage screen.
- 6.9B tokens processed
- 83,022 agent runs in the execution history
- 83.6K tokens per run on average
- ~$156 consumed
His quote from the post:
"Inference with @CommandCodeAI Rocks. 16 windows going every day. Still have plenty of spend left after 3 weeks."
What he shipped
The numbers are only half the story. Here's what those 83,022 runs turned into, in his words:
- Refactored a .NET site
- 2 C++ projects
- 4 new websites
- Refactored 40 cPanel sites
- A Roblox game
- Module for @bridgemindai (soon)
- App for @NousResearch (soon)
That's a .NET refactor, two C++ projects, four new sites, forty cPanel refactors, and a game, alongside two more builds in flight, all run through the same harness in three weeks.
Why this matters
JeffersonNunn.eth isn't running a comparison or making a switch announcement. He's running real work, a lot of it, in parallel, and the spend held up over three weeks of daily use.
Sixteen windows every day for three weeks at 6.9B tokens and roughly $156 is the kind of throughput-per-dollar that's hard to argue with. When a developer keeps that many sessions live and still has budget left, the harness isn't the bottleneck, the day is.
Try it
1npm i -g command-codeRun cmd in your project, open as many windows as your work needs, and watch your usage screen.
Sign up for Command Code · Read the docs · See the original post
