T3 Code · Performance over time
Process CPU per served agent
All metrics
45 / 120
Process CPU per served agent (lower is better)
0.10% per agent
18 Nov−13.4% vs the median of up to 7 previous complete nightlies, within noiseFork perf.9 0.34%, +234.0% vs its upstream baseOnce per record, at start-up: the agents control does not apply.
DatabaseRange
When upstream merges a fix, the lines meet.
Upstream nightlyPerformance forkNoise bandRange of attempts green = better, red = worse, beyond the noise band
Once per record.
Slope of process CPU against the agents served, fitted over the 50 to 250 agent levels of one record: each level's median process CPU against its median agents served (sessions with provider output in the window), which a saturated level keeps below the agents it offers. Null when fewer than two distinct served counts remain.
Lower is better. Noise band 20.7% on Linux x64 (n = 3), so a change counts beyond 25.7%.
- Unit
- % per agent
- Better
- lower
- Noise band
- 20.7%, threshold 25.7% (n = 3)
- Recorded from
- 17 Oct 2026; before that the series is a gap, never a zero
- Platforms
- Linux x64
- Lane
- Every record, inside the agent ladder
- Upstream builds
- Every upstream build
- Source
merge:least-squares slope of proc_cpu_pct medians vs agents_served medians, levels 50–250
Process CPU per served agent (lower is better)
Measured once per record, outside the agent ladder: on the large fixture after its migration, and on an empty database where the catalogue says so.
Records, once per record
| Date | Tag | Median | Min to max | Valid | vs up to 7 previous | Verdict |
|---|---|---|---|---|---|---|
| 18 Nov 2026 | v0.1.11-nightly.20261118.4903 | 0.10% | 0.10 to 0.10 | 1 | −13.4% | within noise |
| 17 Nov 2026 | v0.1.11-nightly.20261117.4851 | 0.11% | 0.11 to 0.11 | 1 | −26.9% | improved beyond noise |
| 16 Nov 2026 | v0.1.10-nightly.20261116.4800 | 0.10% | 0.10 to 0.10 | 1 | −42.6% | improved beyond noise |
| 15 Nov 2026 | v0.1.10-nightly.20261115.4749 | 0.12% | 0.12 to 0.12 | 1 | −44.4% | improved beyond noise |
| 14 Nov 2026 | v0.1.10-nightly.20261114.4709 | 0.095% | 0.095 to 0.095 | 1 | −56.8% | improved beyond noise |
| 13 Nov 2026 | v0.1.10-nightly.20261113.4656 | 0.15% | 0.15 to 0.15 | 1 | −35.8% | improved beyond noise |
| 11 Nov 2026 | v0.1.10-nightly.20261111.4544 | 0.49% | 0.49 to 0.49 | 1 | +116.4% | regressed beyond noise |
| 10 Nov 2026 | v0.1.9-nightly.20261110.4491 | 0.99% | 0.99 to 0.99 | 1 | +338.5% | regressed beyond noise |
| 9 Nov 2026 | v0.1.9-nightly.20261109.4443 | 1.14% | 1.14 to 1.14 | 1 | +404.0% | regressed beyond noise |
| 8 Nov 2026 | v0.1.8-nightly.20261108.4380 | 0.56% | 0.56 to 0.56 | 1 | +149.1% | regressed beyond noise |
| 7 Nov 2026 | v0.1.8-nightly.20261107.4340 | 0.59% | 0.59 to 0.59 | 1 | +160.2% | regressed beyond noise |
| 6 Nov 2026 | v0.1.8-nightly.20261106.4288 | 1.04% | 1.04 to 1.04 | 1 | +361.1% | regressed beyond noise |
| 5 Nov 2026 | v0.1.8-nightly.20261105.4234 | 0.23% | 0.23 to 0.23 | 1 | +7.9% | within noise, thin reference |
| 4 Nov 2026 | v0.1.8-nightly.20261104.4192 | 0.21% | 0.21 to 0.21 | 1 | −5.3% | within noise, thin reference |
| 3 Nov 2026 | v0.1.7-nightly.20261103.4138 | 0.23% | 0.23 to 0.23 | 1 | no reference yet | |
| 2 Nov 2026 | v0.1.7-nightly.20261102.4093 | 1.14% | 1.14 to 1.14 | 1 | no reference yet | |
| 1 Nov 2026 | v0.1.7-nightly.20261101.4048 | 1.15% | 1.15 to 1.15 | 1 | no reference yet | |
| 31 Oct 2026 | v0.1.7-nightly.20261031.4004 | 1.16% | 1.16 to 1.16 | 1 | no reference yet | |
| 30 Oct 2026 | v0.1.6-nightly.20261030.3940 | 1.25% | 1.25 to 1.25 | 1 | no reference yet |
Download this series
series/linux-x64/cpu_per_session_pct.latest.jsonlast 90 days, JSON, 8.1 KBseries/linux-x64/cpu_per_session_pct.all.jsonevery record, JSON, 8.1 KB