RISK — PROCEED ONLY WITH MONITORING
The micro-trauma method had been sitting in the Queue since month one.
The System had labeled it theoretical. Which was its way of saying: plausible, unverified, proceed at your own risk.
The idea itself wasn't complicated. Push the mana channels slightly past capacity. Cause controlled micro-damage. Then heal it with enough precision that the channel rebuilt a little wider than before. Repeat.
It was the same logic as progressive overload. In his first life, that had meant muscle—tearing fibers under load, letting them knit back stronger. Now it was mana channels. Different material, same principle.
He'd left it untouched for four months.
Not because he doubted it would work. He was fairly sure it would. But the margin for error was narrow, and he knew it. Too much stress and the damage stopped being clean. Too little precision in the healing and the channels would scar instead of expand.
Scar tissue didn't scale. It just capped you.
So he waited until the conditions were actually right, not just close enough.
Month five.
Healing precision: 0.4 mm
Channel sensing: fiber-level resolution
Status: Dependency cleared
The risk didn't disappear. It never fully would. But it dropped into a range he could work inside, and that was enough. He started.
Sessions ran forty minutes. The Archive, after hours, when the building was quiet.
Halven's monitoring bracelet stayed on his wrist the whole time. She'd handed it over about two weeks after they met—told him if he was going to run protocols on himself, she wanted the data when something went wrong. Not if. When.
He hadn't argued. It was a reasonable position.
Each session followed the same structure. Twelve mana total. 10.5 pushed into controlled stress—right up to the edge of channel instability, not past it. The remaining 0.5 held back for healing, used only after, precise and deliberate. No overflow. No improvisation. No convincing himself that a small adjustment mid-session was fine.
Then twenty-four hours of rest before the next one.
The Log tracked everything else. Morning readings, evening readings, trend lines over time. It kept him honest. There was no way to talk himself into thinking progress was happening if the numbers said otherwise, and no way to miss it when something actually changed.
Week one: 12 → 14. Slow, but expected.
Week three: 14 → 17. Better.
Week six: 17 → 23.
He looked at that last entry for a while. The jump was too large for the pattern to still be linear. He ran the numbers a second time to make sure he hadn't logged something wrong.
He hadn't.
What was happening made sense, once he thought it through. Wider channels could handle more stress. More stress meant more expansion per cycle. More expansion meant even higher capacity next cycle, which meant more stress tolerance, which meant—
Loop.
The System flagged it before he finished the thought:
POSITIVE COMPOUNDING — CAPACITY TRAINING
Rate: non-linear. Projection advised.
He ran the projection.
Month eight: ~30 mana.
Month ten: ~40 mana.
High variance. Rough estimates. But they were grounded in real data, not guesswork, and that made them worth taking seriously.
Week four brought something else. A footnote in Errevane's paper—a technique called mana saturation. Overpressure method. Faster expansion rates, significantly higher ceiling. Also significantly higher risk.
The System logged it the moment he finished reading:
FUTURE TECHNIQUE — HIGH RISK
Requirement: Higher baseline
He marked it, filed it, and moved on. Queue order existed because he'd learned what happened when you skipped steps. It wasn't efficiency. It was just finding out where you'd cut corners, slightly ahead of schedule.
Month six. 23 mana.
Five Overclock components chained. Full stack duration: twelve minutes.
The System updated its records. Standard. Expected. Then it added something he hadn't asked for.
SYSTEM — SELF-IMPROVEMENT
Pattern recognition: +22%
Cross-reference latency: −18%
Cause: not direct training
Source: accumulated data quality and volume
He read it twice.
The System was getting better at its job—not because he'd done anything to improve it directly, but because he'd been feeding it clean, consistent data over time. Use generated data. Data built patterns. Patterns improved processing. Better processing made the next cycle of use more effective.
Same underlying structure as the channels.
He hadn't planned that. He wasn't sure it was plannable. But it was happening, and it mattered.
He flagged it.
HIGH SIGNIFICANCE
And left it there, for when he had time to think through what it actually meant.