Experience and methodology
Experience, measured carefully.
Zero Width Lab states what each credibility figure covers, where it came from, and what it does not establish.
AI tokens logged in hands-on development
This local snapshot aggregates numeric token counters available across two Claude Code profiles and Codex session records through 1 September 2026. Counted records span 11 June through 1 September 2026, a period of less than six months.
- Claude Code logs
- 17.14 billion
- Codex session logs
- 17.84 billion
- Reused working context
- about 95.8%
What the number means
AI tools record tokens whenever they process text during a working session. This total adds the available numeric counters from two Claude Code profiles and Codex session records. Missing categories were not estimated.
Most of the total comes from working material the tools reused from earlier turns. Those tokens still appear in the logs, but they are not unique new text. The figure shows the volume of hands-on AI use—not customers, projects, production deployments, or successful outcomes.
35+ years of technology experience
The technical practice behind Zero Width Lab draws on at least 35 years of hands-on information-technology experience, based on the owner’s attested practitioner history.
This is a statement about practitioner experience. It does not mean that Zero Width Lab operated as a company throughout that period.
12+ implemented projects
A privacy-safe local inventory verified at least 12 distinct implemented software, automation, tooling, and research projects.
The count requires a distinct objective and material deliverable:
- software or product applications;
- automation and technical tooling; or
- implemented research and experimental systems.
The count excludes planning-only folders, trivial scripts, forks, worktrees, repeated product copies, and supporting components counted as separate projects. It does not represent clients, commercial engagements, or production deployments.
Independence and privacy
Zero Width Lab independently uses Claude Code and Codex. Anthropic and OpenAI have not audited, sponsored, certified, or endorsed this analysis.
The public aggregate contains no account identifiers, prompt or response text, repository paths, credentials, or client content. Local session logs are not provider-audited billing statements, and tools may define token categories differently.