The Filing
He was at his desk by ten-fifteen. The flight had come in at nine-oh-four, the taxi had taken forty minutes through the Wednesday evening traffic, and he had eaten nothing since the airport sandwich in Tanger, which he had eaten standing up at a counter without tasting it. He made coffee, opened his laptop, and spread Yara's documents to the left of the keyboard and his own notebook to the right, and he looked at the shape of the thing before he began.
He had twenty-six hours to the consultation close. He needed six to write, one to review, which was not enough but was what was available. He had done more with less.
He opened a new document and began.
The lede came quickly, which was a good sign. The lede was always the diagnostic — if it resisted, the story had not yet resolved into the shape it needed, some structural problem still alive in the material that the writing would eventually surface. This one arrived in two sentences and he left it there and moved into the body and the body moved.
The architecture of the piece was Yara's architecture, because Yara's architecture was the truth of the thing. The v2.4 specification first — what it was designed to do, how it worked, what the audit trail looked like under it and what it did not contain. He rendered it in the same precise, almost-plain prose that the specification itself used, because the specification had made the choice that clarity was a tool, and he respected this, and because the technical literacy his editor had told him was an advantage was the literacy that allowed him to follow it. He did not simplify. He trusted the reader to follow.
The accountability attribution memo next. The legal department's finding that a system built to v2.4 specifications would satisfy the consultation's human oversight requirement and simultaneously make that oversight unauditable. He quoted from the memo in two places — the phrase regulatory alignment, used three times by the legal team as a term of art — and on each occasion he let the phrase sit without comment, because it did not need comment. A phrase that had been chosen by lawyers to describe the design was the best evidence of the design's intent.
The continuous monitoring framework last in this section, and he spent the most time here, because this was the inversion that required the most careful scaffolding. Hartog had presented it at the board presentation as the technically rigorous alternative to the consultation's current framework, and his argument was sophisticated, and he had been right that the consultation's framework had problems, and none of that was in tension with the specification's own account of what continuous monitoring did and did not measure. He was careful with this. He was fair to Hartog's argument, stated it clearly, acknowledged its merits, and then placed beside it the specification's own language: system-level behavioural consistency, not individual decision auditability. He was not writing an opinion piece. He was placing two things next to each other and allowing the reader to complete the geometry.
He wrote this section twice. The first version was six hundred words and was correct and was also, he felt, subtly prosecutorial in its structure — the sequence of revelation, the rhythm that pointed. He deleted it and wrote it again. The second version was nine hundred words and moved like the Crichton document it was drawing on: expert-knowledge as disclosure rather than argument, the reader discovering as Daniel had discovered, the prose tracking understanding in real time rather than presenting its conclusions in advance. He read it back and it was better and he left it.
He reached the sourcing section at one in the morning and stood up and made more coffee and looked out the window at the Amstel, which was dark and reflecting the bridge lights in long orange columns, and he thought about what he was going to do with the board minute.
The board minute was his original document, the one in the package, the one with the 412,000 figure. He needed it in the piece — it was corroborating evidence for Yara's specification, confirmation that the v2.4 architecture had been presented to and approved by the Arken board at the level the specification claimed. Without it, the piece rested entirely on Yara's documents and Yara's testimony, which was strong but which Arken would characterise as a single disgruntled former employee. With it, there were two independent document trails, and the board minute provided a date, a formal record, and a named list of executives present.
He needed it. He would use it. He would cite the daily decision volume from Arken's own public technical disclosure — 312,000, the correct figure, four clicks from anyone's browser — and not from the board minute, and he would not explain why he was doing this, because the explanation required sourcing he did not have.
This was the first reason.
He could not prove the 412,000 figure had been placed in the document deliberately. What he had was a figure in his document that did not appear in any of Yara's documents, that did not appear in Arken's public disclosures, that matched the figure in a retracted Brussels piece from six months ago, and a phone call from Hartog in which the Brussels retraction had been raised and withdrawn in the same breath with the specific quality of a warning that could not be called a warning. He had a pattern. He did not have a source, a chain of custody, a documentary record of how the figure entered his document. He had three conversations with three people and a forty-eight-hour window during which the document had been delivered to his flat by hand. He could not prove intent. He could not even prove a connection between the Brussels retraction and his document that would survive a lawyer's review.
Without proof it was not publishable. This was the second reason, and it was the same as the first reason stated from a different direction, and he acknowledged this without treating it as a problem, because it was the condition of the profession and always had been.
The story was true regardless. This was the third reason and it was different from the first two in a way that mattered. He had Yara's documents. He had verified them against public record and against the board minute and against Arken's own filings. The v2.4 specification was real. The accountability attribution gap was real. The continuous monitoring framework as companion rather than replacement was real. The story he was writing — the architecture of what Hartog had built, and what it would do to the regulatory framework if the consultation closed on the current language — was true. Publishing it was the correct thing to do with true information before a regulatory window closed. Not publishing it was the other kind of cost.
Yara had given this to Thierry Marchand in September. Thierry had filed, and the story had been retracted, and the consultation had resumed, and no other outlet had picked it up.
He stood at the window for a moment longer, the Amstel flat and orange-lit below him, and then he sat back down and kept writing.
The piece finished at ten fifty-eight. He read it through once without changing anything — a single read, for shape and proportion — and then made twelve edits, mostly in the final third, where the prose had tightened under tiredness into something that was more efficient than it should be. He read it again. It was good. It was the best piece he had written on this beat in two years, possibly longer, and it was built on documents that were real and a source whose technical knowledge was beyond his ability to challenge and whose motivation he understood. He believed what it said.
He sent it to Petra at eleven-oh-four, with a note that was four sentences: the lede was slightly slow through the second paragraph but he had not found the fix; the Hartog direct quote from the board presentation was confirmed against his recording; Yara's documents were available to the legal team and she had agreed in principle to provide a written statement; he was available for editorial queries until midnight.
He closed the laptop.
He sat in the silence of the flat for a moment — the clock on the kitchen wall, the bridge lights on the water — and then his phone moved on the desk. A message from Petra.
He looked at the timestamp.
She had replied at eleven-oh-eight.
He sat with the phone in his hand. Four minutes. He had told her he was filing by eleven. She had said she would be there. She could have been at her desk, the piece open as it arrived, reading as it downloaded, ready. This was possible. This was the innocent version.
He picked up the notebook. He opened it to the last written page and below the last line he wrote four numbers.
11:04. 11:08.
He looked at them for a moment. Then he closed the notebook and placed it on the desk beside the phone and did not pick either of them up again.
The consultation closed in twenty-five hours. He had done what he could do with what he had. He turned off the desk lamp and the flat went dark except for the bridge lights on the water, which did not change, and he sat there in the dark for a while, not moving, thinking about nothing in particular, before he went to bed.