Sunita

The Priority 1 case was eleven days old.

Arjun was aware of this the way he was aware of most things he wasn't acting on — as a peripheral fact, present and untouched. He had read the full documentation twice now. The confidence interval. The sixty-three pages of ecosystem data, projections, relocation feasibility. He had read the three-page description of the Mawali community and noted the word moderate. He had looked up the definition of moderate. He had not opened a new document. He had not moved the cursor back into the signature field.

He was reviewing a case file when Sunita stopped at the edge of his desk.

'You're on the wetlands,' she said.

It wasn't a question. The Priority 1 tag was visible on his screen from where she was standing.

'Among other things,' he said.

She had a coffee in one hand and her tablet in the other. She had already signed her morning queue — he'd watched her do it from across the open office floor, the quality of attention she brought to it, which was complete and efficient and entirely without the hesitation he could never seem to shed. Three cases in just under forty minutes. He had timed it without deciding to.

'How do you do it?' he said.

She looked at him.

'Sign that fast,' he said. 'Without — ' He stopped. He wasn't sure how to finish the sentence without it sounding like a criticism of her or a complaint about himself, and he didn't intend either.

Sunita considered this for a moment. She wasn't offended. She set her tablet on the corner of his desk and looked at the wall screen where the queue updated in real time, the numbers rolling.

'The model is more accurate than I am,' she said. 'In cases where the confidence interval is above ninety percent, the model is right more often than I would be if I second-guessed it. That's just the data.' She picked up her tablet. 'My job is to be present when it's wrong. If I spend my energy on the cases where it's right, I won't have anything left for the ones where it aren't.'

She said this without self-pity, without irony, without any of the defensive register he might have expected. She meant it. It was a coherent position, worked out and settled. She had arrived at it the way people arrive at positions they've had to defend enough times to stop needing to defend.

'Your wetlands case is below threshold,' she said. 'Eighty-seven point three.'

'Yes.'

'That's why it's yours.' She picked up her coffee. 'That's why it's not mine.'

She went back to her desk.

Arjun looked at the Priority 1 file on his screen. He looked at it for a moment, then opened it.

He read the executive summary again. He read the confidence interval breakdown — the 12.7% of uncertainty distributed across ecosystem resilience variance and stakeholder response modelling. He read the relocation feasibility assessment for the third time. Subsistence-dependent, culturally attached, relocation timeline: 24 months, disruption level: moderate. On the second page of the assessment there was a table of comparable relocation programmes — six cases, the Habitat Restoration Authority's previous implementations under similar conditions. Mean recovery time: four point two years. All six within the five-year expected recovery window. The data was internally consistent.

He closed the relocation assessment. He opened the ecosystem resilience projections. The model's recommendation was not close. It wasn't a marginal call. The western buffer zone was losing viability at a rate the projections showed clearly, and the freshwater diversion capacity required to reverse that loss would have to come from the core system — from the zones where the Sundari had actually survived, where the recovery was real and measurable and ongoing. The trade-off was documented. It wasn't the model's opinion. It was what the data showed.

He knew this. He had known this after the first reading.

He closed the projection document. He looked at the signature field.

He got up. He walked to the window at the end of the office floor. Dhaka below: the elevated ring, the construction, the density of the mid-city pressing outward in all directions. The towers to the north were new since he'd started here — two of them finished, one still in frame, the upper floors wrapped in construction mesh. The population that had moved from the coastal zones in the last twenty years was somewhere in there, in the apartments and the infrastructure and the new mass transit lines that the city had built to absorb them. The delta had sent its people here. The city had taken them. The adjustment was ongoing and largely invisible from this window.

He went back to his desk.

His monthly allocation summary was open in a background tab. He had looked at it that morning — routine, the way he checked it at the start of each month. Twelve cases in the active queue. He looked again now. Three Priority 1 items. He was certain there had been two this morning.

He checked the time stamp on the new item. It had been added forty minutes ago. A mangrove boundary reclassification in the northern reach, confidence interval 84.1%, stakeholder complexity: high. He opened it and read the first page.

He could feel the shape of the afternoon assembling: the new Priority 1, the wetlands case still unsigned, the annotation he hadn't yet written for either of them. He looked at the cases in the queue, the way they had been arriving, the particular weight of what came to him compared to what he heard from colleagues in the adjacent bays. Sunita's morning. Elena's queue, which she cleared in two days on a consistent cycle and which he had never seen carry a Priority 1 item.

He noticed this. He let it sit.

He opened the wetlands case one more time. He looked at the confidence interval. He thought about what Sunita had said — not its conclusion but its logic, the way the position was structured. Present when it's wrong. It was a form of triage. A decision about where to put the weight of attention. He did not disagree with it as a principle. He understood why it worked for her.

He thought about the Mawali community. The table of comparable relocations. Mean recovery time: four point two years. He had no reason to believe this case would fall outside the documented range. He had no data that contradicted the model's assessment. He had only the site visit, six months ago, and what he'd seen at the edge of Banshtala in the early morning — the collectors, the equipment check, the way the older woman had read the canopy — which he had filed nowhere and which was not a data point.

He opened a new document.

He sat for a moment with the cursor in the blank field.

Then he closed the document without writing anything. He moved to the new Priority 1 item — the boundary reclassification — and began to read.

On the wall screen across the office, the queue updated. The Priority 1 tag on the wetlands case had been there for eleven days. It would be there tomorrow. The deadline was March 31. There was time.

Outside, a thin rain had started against the windows, the kind that barely registered — just a quality of the glass, a greying of the city beyond it, the towers going slightly soft at the edges.