Codex activity ledger
Task timestamps, token totals, goal runtime, and tool events
Typed fixture only in this prototype; no private account data is read.Account boundary: Ronova IDDefinitions and unreasonable care
Methodology
The registry is funny only if its measurements remain honest. Each family declares the period, threshold, source, uncertainty, and correction path before a badge is issued.
Registry principles
Precision where possible. Ranges where necessary.
- 01
Define before counting
Thresholds, time windows, and units are versioned before a user’s value is compared.
- 02
Expose the source
Every detail page states the intended evidence source. This prototype labels fixtures rather than implying private live access.
- 03
Keep uncertainty visible
Estimated values display ranges, assumptions, and confidence instead of decorative decimal places.
- 04
Permit correction
An erroneous distinction can be corrected or revoked; the reason remains in the evidence timeline.
Metric definitions
Six initial departments
- Tokenburner
Sum reconciled input and output tokens across eligible coding tasks. Cache semantics and deleted history are disclosed as evidence limitations.
Local Codex activity export and reconciled task ledger- Limit Killer
Compare the first reset marker with the first subsequent provider limit response; shorter is regrettably higher.
Reset markers and local limit-response timestamps- Spark!
Sort timestamped task usage, slide a 24-hour interval across the ledger, and retain the maximum inclusive sum.
Local timestamped task ledger- Marathon With No Ending
Use goal lifecycle timestamps and subtract only deliberate pauses. Tool latency remains part of the institutional experience.
Goal lifecycle events- SSD
Inventory one workspace on disk without following external symlinks; report source separately from total footprint.
Local filesystem inventory- Sustainable Person
Combine low/high estimates for model compute, local device energy, and grid intensity. Never treat the result as provider-certified emissions.
Activity estimates and published energy-intensity rangesThe displayed 180–340 kg CO₂e interval is illustrative and carries medium confidence.
Approximate impact office
180–340 kg CO₂e
- Estimated energy
- 310–610 kWh
- Assumed hardware
- Mixed local Apple silicon and provider accelerator workloads
- Grid assumption
- Illustrative 0.31–0.56 kg CO₂e/kWh operating interval
- Provider uncertainty
- Provider allocation, cooling overhead, cache treatment, and embodied emissions are not independently available.
- Confidence
- medium
- The interval is intentionally wide and should not be read as a provider emissions statement.
- Location is represented as a grid-intensity range rather than one false-precision coordinate.
- Corrections replace prior estimates; the registry retains a visible correction note.
Data-source status
Nothing connected without disclosure
Workspace inventory
Repository size, generated output, caches, and dependency directories
A real connector would require explicit local permission and a disclosed scan root.Account boundary: Ronova IDGit history
Commits, branches, merges, rollbacks, and CI evidence
Unavailable in this non-Git checkout; no result is fabricated.Account boundary: Ronova IDInstitutional boundary
Public fixture. Serious methodology. One account system.
Archivements currently uses typed sample records. Any future private controls or connected sources remain behind the universal Ronova ID; this project does not create another login, session, or identity database.