Numbers
Every figure on this site, grouped by where it comes from, with its source and date.
01
| Figure | What it measures | Source | Date |
|---|---|---|---|
| 76 | development tickets tested and approved before release (103 approval events) | Own records · current role | 2026-10 |
| 18 | tickets sent back to development because they did not pass | Own records · current role | 2026-10 |
| 30 | tickets checked again in production after release | Own records · current role | 2026-10 |
| 171 | tickets in the board history behind these figures | Own records · current role | 2026-10 |
| 13 to 45 | cases per end-to-end test matrix (the largest: 45, 40 and 36) | Own records · current role | 2026-10 |
| 27 | documented test-case folders, so a release can be re-tested the same way | Own records · current role | 2026-10 |
| 22/22 | payment-receipt cases passing after a billing fix | Own records · current role | 2026-10 |
| 6 in < 0.9 s | back-to-back customer searches on an account with about 30,000 customers, with no restart | Own records · current role | 2026-10 |
| 30+ | QA verdicts logged in one week (6–12 Oct 2026) | Own records · current role | 2026-10 |
02
| Figure | What it measures | Source | Date |
|---|---|---|---|
| 50+ | customer branches whose data I moved from Excel or older systems into the live platform | Own records · current role | 2026-10-13 |
| 8,339, 0 errors | past orders from 3 branches migrated and reconciled against the source files | Own records · current role | 2026-09 |
| 18 files, 6,657 rows | analyzed for one branch before any migration code was written | Own records · current role | 2026-09 |
| 188 | orders repaired after a field the app requires was found empty | Own records · current role | 2026-09 |
| 1,101 + 126 | labor concepts and service packages cloned to 5 branches | Own records · current role | 2026-09 |
| 5,908 | spare parts loaded into 24 customer accounts | Own records · current role | 2026-07 |
| 1,485 | purchase groups rebuilt after the first load had 3 errors | Own records · current role | 2026-07 |
| 58 of 59 | rows mapped in a rehearsal run that does not touch production, 0 errors | Own records · current role | 2026-07 |
| ~3,400 vs 11,934 | packages a rehearsal run counted vs the real total; the gap exposed unordered pagination | Own records · current role | 2026-06 |
| ~21,539 + ~11,934 | labor concepts and service packages corrected in one controlled bulk update | Own records · current role | 2026-06 |
03
| Figure | What it measures | Source | Date |
|---|---|---|---|
| 300+ | active customer accounts on the platform | Own records · current role | 2026-10-13 |
| 10+ / 100+ | on-site implementations / remote rollouts | Own records | 2026-10-13 |
| 6,000+ / ~60 h | CRM contacts / hours of manual work removed each month by the WhatsApp automation | Own records · current role | 2026-10-13 |
| 22 → 11 | customers with late payments contacted → accounts brought up to date (1 closed) | Own records · current role | 2026-09 |
04
| Figure | What it measures | Source | Date |
|---|---|---|---|
| 71 of 243 | commits co-authored by AI in my work repository | Own records · current role | 2026-10 |
| 76 | Claude Code work sessions logged automatically by a hook since 28 Sep 2026 | Own records · current role | 2026-10 |
| 3 | agent roles, each pinned to a model by difficulty and cost: plan, execute, explore | Agent definitions | 2026-10-14 |
| 9 | sections every plan must have before an agent executes it | Planning rules | 2026-10-14 |
05
| Figure | What it measures | Source | Date |
|---|---|---|---|
| up to 40 | people in the largest team I led (IPN) | CV | 2026-10-13 |
06
| Figure | What it measures | Source | Date |
|---|---|---|---|
| 137 | data models | Counted in the repository | 2026-10-13 |
| 96 | versioned database migrations | Counted in the repository | 2026-10-13 |
| 107 | end-to-end smoke scripts against the real API | Counted in the repository | 2026-10-13 |
| 722 | REST route decorators in the API | Counted in the repository | 2026-10-13 |
| 117 | pages in the web app | Counted in the repository | 2026-10-13 |
| 125 | tables isolated per clinic with row-level security | VetNova status file | 2026-09-19 |
| 8,597 | translation keys in 158 namespaces | VetNova status file | 2026-09-19 |
| 5 | interface languages | VetNova status file | 2026-09-19 |
| 18 of 54 | ad clicks with real search intent; 32 (59.3%) were Display clicks | VetNova decision log | 2026-09-21 |
07
| Figure | What it measures | Source | Date |
|---|---|---|---|
| 10,000,000 | synthetic policies in the data-model benchmark | REYBER prototype report | 2026-09-22 |
| 492 vs 3,030 MB | search index of the chosen design vs the alternative (6x smaller) | REYBER prototype report | 2026-09-22 |
| 1.2 vs 118 ms | two-attribute filter, chosen design vs the alternative | REYBER prototype report | 2026-09-22 |
| 233 s, live | to make a new attribute searchable with the portfolio online (113 s with it locked) | REYBER prototype report | 2026-09-22 |
| 25 | research and requirements documents before the first feature | Counted in the repository | 2026-10-13 |
| 70 | tables with a tenant, all under row-level security | REYBER status file | 2026-10-13 |
08
| Figure | What it measures | Source | Date |
|---|---|---|---|
| 13 | request types answered by plain code, without the AI model | JARVIS status file | 2026-10-13 |
| 47/47 | router tests passing | JARVIS status file | 2026-10-13 |
| 2.7 s | fast-path answer (target 3.5 s, met) | JARVIS status file | 2026-10-13 |
| 8.6 s | answer when the model responds (target 6.0 s, missed) | JARVIS status file | 2026-10-13 |
| 8/8 | brain test with the real model, safety cases included | JARVIS status file | 2026-10-13 |
09
| Figure | What it measures | Source | Date |
|---|---|---|---|
| 57 | songs in the lyric-video catalog | Project status file | 2026-10-13 |
| 19 | coloring-book volumes generated (1 published) | Project status file | 2026-10-13 |
| 60 | items in the live jewelry catalog | Live site, checked | 2026-10-13 |
| ~35% | fewer tokens per video analysis | Project rules file | 2026-10-08 |
| ~70× | faster image generation on a rented GPU than on the laptop | Own measurement | 2026-10-13 |