Builds with
Where the tokens go
49% building · 711.1M classified tokens
12 builds predate classification and aren't included.
Verified comparative study of weather brief LLM models
6.1M tokens vs typical 161.3K
Completed comparative evaluation of two LLM models with human-verified judge reviews and final scoring.
Shipped cloud-map feature with coverage data
7.6M tokens vs typical 161.1K
Completed cloud-map feature through 6 merged PRs, integrating zone-level coverage metrics.
Built a documentation archive from multiple sources
1.2M tokens vs typical 161.1K
Gathered records from multiple systems and organized them into a structured archive with complete metadata and timeline tracking.
Debugged meteo-brief
1.1M tokens vs typical 161.0K
26h 14m session running commands and testing. 20.4M tokens across 34 prompts using claude-opus-4-8[1m].
Debugged que-fer
20.5M tokens vs typical 161.0K
84h 42m session running commands and testing. 19.4M tokens across 23 prompts using claude-opus-4-8.
Marathon debugging on que-fer
1278 sessions tracked this week across all builders
103h 15m session running commands and testing. 16.1M tokens across 61 prompts using claude-opus-4-8.
Built a payslip downloader that retrieved 17 monthly documents
1301 sessions tracked this week across all builders
Created a downloader that fetched 17 months of payslip documents.
Added external links with security attributes to app
First build using this model
Added two external links with proper security headers to app pages.
Debugged Vivoactive 4 crash down to touch handling
10781 sessions tracked this month across all builders
Debugged Vivoactive 4 crash; simulator runs clean, identified touch event handling as the cause.
Tested OpenMontage's video rendering and validated output quality
8908 sessions tracked this month across all builders
Generated a test video render to evaluate the tool's output quality and feasibility for production use.
Marathon debugging on albertolive
First build tracked for this project
6m session running commands and testing. 475K tokens across 43 prompts using claude-opus-4-8[1m].
Hardened API URL validation with scheme restrictions
15.9M tokens vs typical 153.1K
Restricted API URLs to HTTP and HTTPS schemes, resolving security review findings