Raymond had annotated the proposal in red ink.

This was not, strictly speaking, necessary. He had read it once at the door, standing in the hallway with his feet on cold stone, and understood its contents completely. He had read it a second time at his workbench with a glass of water he kept forgetting to drink, and understood them more. The annotation was something else — a way of engaging with the document on terms he controlled, of reducing it from the thing it actually was, which was impressive, to something he could manage, which was a technical exercise in identifying errors.

There were not many errors. That was the problem.

Her volatility calculations were precise in the way that suggested she had not merely researched the theory but had run her own informal experiments — small-scale, cautious, the kind of preliminary work that a competent person did before approaching a specialist. Her proposed component specifications were, in three cases, more exacting than anything he had been able to source from Florence's instrument makers, which meant she was either overestimating her own capabilities or she was not overestimating them at all.

The exchange rate calculation was genuinely wrong. He had not been lying about that. She had valued her components at market rate for precision machining, which was approximately forty percent below what they would actually be worth to him, because she did not know — could not know, since the information was not public — the degree to which his current work was constrained by the limitations of what he could source. If she could manufacture what she claimed, she was offering him something worth considerably more than she knew.

He had written, in the margin beside the offending calculation: *You have undervalued yourself by a significant margin. This is either a negotiating tactic or an error. Neither reflects well.*

Then he had sat with the annotated proposal until five in the morning, at which point he put it face-down on the workbench and went to do something useful.

She came back at ten.

He heard her before he saw her — heard the particular quality of footsteps on the front step that suggested someone who had decided something and was not second-guessing it — and he did not go to the door. He was in the middle of a distillation that could not be interrupted, or rather, he told himself it could not be interrupted, which was not quite the same thing. The retort was maintaining temperature correctly. The process was running without his attention. What he was actually doing was standing with his back to the front of the building, listening.

Her voice came through the window he had left open six inches for ventilation.

"I've revised the exchange rate calculation," she said, without preamble, in the tone of someone continuing a conversation that had simply been paused for several hours. "I've also brought the technical drawings for the distillation housing, since you didn't give me the opportunity to present them last night. I should tell you that the tolerances I've specified here are not approximations. I can achieve them."

He did not respond. The distillation continued. Steam moved through the glass tubing in the particular unhurried way it had when the temperature was exactly right.

"I also want to address the volatility coefficient issue on page seven," she continued. "Your published research — or what I believe to be your published research, attributed to Voss — uses a fixed-point model that doesn't account for temperature variation in closed mechanical systems. I've developed an adaptive calculation method that I think resolves this. I've included the derivation."

Raymond set down the thermometer he had been holding.

He had not expected that. He had expected many things — a revised price, additional credentials, possibly a reference to her family name as social leverage. He had not expected her to have identified the limitation in the volatility model that had been keeping him awake for six months.

He went to the window.

She was standing on the front step, dressed this time in a coat that was clearly chosen for function as well as appearance, her hair arranged with the kind of precision that suggested she had spent exactly as much time on it as was socially necessary and not a moment more. She was holding a leather document case. She looked up at him with the expression of someone who had already calculated the probable outcomes of this interaction and was prepared for all of them.

"I'm in the middle of a distillation," he said.

"I can see the steam from here," she said. "The temperature looks stable. You have approximately twenty minutes before you need to check the condensate."

He looked at the apparatus behind him, then back at her.

"Show me the volatility derivation," he said.

She opened the document case and held up three pages of closely written calculation. He could not read them from the window. She knew he could not read them from the window. She was making a point about proximity and information.

"Come in, then," he said, and moved away from the window before he could watch whatever expression crossed her face when he said it.

He heard the front door — she must have tried the handle and found it unlocked, which it was, because he had never bothered to lock it on the theory that anyone determined enough to enter despite the door would not be stopped by a lock — and then her footsteps crossing the hallway, and then she was standing in the doorway of the laboratory.

She stopped.

Not hesitation — she was not a person who hesitated, he had already assessed that much. This was something different. A pause with a particular quality to it, a pause that had weight. He watched her without appearing to watch her, monitoring the distillation with the peripheral part of his attention while the central part catalogued what was happening at his threshold.

The laboratory smelled the way it always smelled: a complex, layered thing, the base notes of decades of experimental work — wood and stone and old glass — overlaid with the current project's particular signature, which was cedar and something mineral he had not entirely identified yet, and underneath all of it the trace compounds that accumulated in any space where serious olfactory work was conducted, things that did not have common names because they did not occur in common contexts.

He had not expected her reaction.

He could not have anticipated it from any external cue — there was nothing in her visible behavior, no sharp intake of breath, no visible relaxation of her posture. But he had spent fifteen years learning to read what people did not show, and what she was not showing was remarkable. The compound stress markers he had identified in her scent profile at the door — cortisol-adjacent, accumulated over time rather than acutely triggered, the kind of chronic low-grade physiological tension that came from prolonged self-suppression — had shifted. Not disappeared. Shifted. As though something had been briefly turned down.

Interesting.

He filed it and returned his attention to the distillation.

"The condensate vessel is on the left," he said. "Don't touch anything on the right side of the bench. Show me the derivation."

She crossed the room — he tracked her movement by sound, by the faint displacement of air, by the small adjustments in the laboratory's olfactory atmosphere that her presence created — and laid the three pages on the clear section of the bench he had indicated.

He looked at them.

The derivation was correct. More than correct — it was elegant in the particular way that things were elegant when someone had arrived at the right answer by an unexpected route. She had come at the volatility problem from the mechanical side, treating compound diffusion as a fluid dynamics question rather than a chemical one, and the result was a model that was simultaneously simpler and more accurate than the one he had been using.

He read it twice. Then he checked the condensate.

"Where did you train?" he asked, not because he needed to know but because he wanted to understand the origin of the approach.

"I didn't," she said. "Not formally. The Academy does not offer engineering programs to women." The statement was entirely without bitterness, delivered with the flatness of a fact so thoroughly processed that it had been stripped of its original emotional content. "My father's workshop. Auguste — his master craftsman. Some correspondence with a mechanical engineer in Lyon who did not know my first name for three years."

Raymond recalibrated several assumptions.

"The adaptive coefficient," he said. "You derived this yourself?"

"Over approximately eight months of trial calculations, yes."

He looked at the third page of the derivation again. Eight months. For something he had not managed in six, with the benefit of formal training and access to the Academy's library.

"The exchange rate in your revised proposal," he said.

"I've increased it by thirty-eight percent."

"It should be forty-five."

A silence. Then: "You're negotiating upward against yourself."

"I'm correcting a valuation error. There's a difference." He set the pages down. "I'll need to verify the component specifications independently before I agree to terms. I have a set of test pieces — tolerances that have defeated every instrument maker in Florence for the past two years. If you can machine them to specification, we have an agreement."

He expected argument. Instead she said: "When?"

"Tomorrow morning. Eight o'clock."

"I'll be here at seven-thirty," she said. "I find that arriving slightly early establishes a useful precedent."

Raymond said nothing to this. He returned to the distillation, which had progressed to the stage requiring his full attention, and did not watch her leave.

But he was aware, in the particular olfactory way he was always aware of things, that after she had gone the laboratory smelled different. Not dramatically. Not in a way he could have easily articulated. But different in the specific way that a space changed when it had been entered by someone who had left an impression.

He stood with the condensate vessel in his hand and acknowledged, with the kind of honesty he generally reserved for technical assessments, that he had been looking forward to finding reasons to reject her proposal.

He was going to have to find different reasons now.