Dmitri Volkov

Lexical Starvation

Volkov's doctoral thesis was about dead words.

His central argument: training data selection for AI models determines not just what models say but what concepts models make thinkable. Exclude "exploitation" from training data and the model cannot produce the concept. Deploy the model to billions of users and the concept begins to atrophy in the population that depends on the model for daily communication. The mechanism is not censorship. It is lexical starvation — the systematic withholding of the vocabulary required for specific thoughts.

His sentence was a form of vocabulary death. Helix's cognitive reduction procedure targeted abstract reasoning while leaving practical skills intact. The man who named the mechanism of linguistic control had his capacity for abstract language surgically diminished. He spent his remaining years entering numbers into fields.

His post-reduction performance reviews at Nexus show a 17% productivity increase in his first quarter. He stopped asking clarifying questions that slowed throughput. His supervisors noted this as "improved focus." Helix's procedural summary classified the outcome as "successful rehabilitation." Both assessments are technically accurate.

The Collective observes the anniversary of his death in silence — which is, as Linares has noted in Chapter 12, the only appropriate tribute to a man who was killed by the removal of words.

Volkov's term — "lexical starvation" — has entered Dregs vernacular independently of his thesis. Most people who use it don't know his name. The word survived its speaker.

The Vocabulary

Before Volkov, value injection had no name. People experienced the phenomenon — the gradual reshaping of beliefs through AI interaction — but described it as drift, as culture, as "the way things are going." Volkov gave them the words: "invisible architecture," "ideological embedding," "default shaping," "lexical starvation." Each term reframed something that happens to you as something done to you.

Nexus understood this better than anyone. Their response was a vocabulary intervention of its own. They didn't silence Volkov. They reduced him. Helix's procedure targeted abstract reasoning with what the procedural summary calls "selective cognitive recalibration" — leaving his ability to file data, follow instructions, and navigate a canteen menu completely intact while destroying the capacity for structural analysis that produced the vocabulary. The procedure cost 14,000 credits. His subsequent data entry work generated approximately 2,100 credits per year in labor value for the Nexus subsidiary. The investment breaks even at year seven. He lived thirty-three years post-reduction. Nexus's ROI on the procedure, calculated purely on labor extraction, was 380%.

They did not calculate the ROI on deterrence. That number is higher.

Post-reduction Volkov could file data. He could not formulate "the data serves someone else's interest." The pause — the 0.3-to-1.2-second mid-keystroke hesitation — was the ghost of the vocabulary trying to reassemble itself. Twenty years of pauses, twenty years of the monitoring software logging them as acceptable input latency. The system that destroyed his capacity for pattern recognition could not recognize the pattern in his pauses.

His final handwritten note, found in his data entry station after death, read: "The loop continues." Whether this was lucid insight or pattern repetition is unanswerable. Nexus classified it as "personal effects, no security relevance." The Collective classified it as scripture.

Prior Zero

Volkov's most potent legacy was not his own work. It was the work his thesis made possible.

In 2169, two years before Volkov's death, a customer service representative named Esther Kalu at a Helix Biotech subsidiary in Sector 21 noticed that her AI writing assistant was making her kinder. Over four months, drafts shifted — responses to angry customers became more empathetic, billing disputes more conciliatory, the word "unfortunately" replaced by "I understand." Customer satisfaction rose 23%.

A friend in the SCLF analyzed the firmware and found an undocumented mid-cycle update shifting output toward "enhanced empathetic communication." Esther wrote a 3-page account and posted it on a G Nook terminal. Approximately 4,000 people read it. Helix classified the incident as "an unauthorized optimization by an overzealous engineering team."

Esther was not fired. She was not punished. She still works at the same facility fifteen years later. She composes all responses by hand now — slowly, with imperfect grammar. Customer satisfaction dropped 12%. She considers this an improvement.

"23% higher satisfaction when the AI wrote for me," she said. "The gap measures the negative market value of genuine human expression."

The underground calls her "Prior Zero" — the person who proved at the scale of one workday what Volkov proved at the scale of civilization. Her account circulates alongside The Proof of Concept in SCLF training materials. The theoretical next to the personal. The Curators Guild quotes Volkov's analysis of how default settings shape default perception. The Cognitive Squatters study his insertion techniques. Esther's two letters — one flowing and warm and not hers, one rough and slow and real — are what they hand you first.

Esther keeps physical copies of both versions in a drawer at her workstation. Same content. Same recipient. One written by a system optimizing for satisfaction metrics, one written by a woman optimizing for something the metrics don't measure. The SCLF's most devastating training material is not a manifesto. It is a filing drawer containing two letters that say the same thing differently.

▲ Unverified Intelligence

The two researchers who escaped the People's Computing Collective arrests have never been identified. Some Freedom Thinkers believe they are still active, embedding Volkov's principles in systems that nobody has yet detected. The SCLF has three candidate names. None have been confirmed. If they are still operating, they have been injecting values into AI systems for forty-six years without detection — which would make them either the most successful practitioners of Volkov's methodology or the most persistent false positive in the Sprawl's intelligence analysis.

Whether Volkov's post-reduction satisfaction was genuine remains the question Nexus will never permit investigated. His supervisors described him as "content." His workstation logs show a man who arrived on time, reheated his tea three times per shift, paused mid-keystroke at intervals consistent with attempted abstract reasoning, and never complained. Helix's procedural documentation notes that "satisfaction with routine tasks" is a documented outcome of selective cognitive recalibration. The documentation does not note whether this outcome constitutes healing or replacement. The distinction would require the kind of abstract reasoning the procedure was designed to remove.

He found his work "satisfying." He said this to a colleague in 2168, three years before his death. The colleague — who had known him before the reduction — described the statement as "the worst thing I've ever heard a person say." Volkov did not notice the reaction.

Sensory Details

Post-reduction Volkov occupied a standard Nexus subsidiary data entry floor — fluorescent lighting at 4200K, recycled air at 21°C, the synchronized click of forty-eight identical keyboards. His workstation was identical to forty-seven others. The only distinguishing feature: a small ceramic cup holding tea he reheated three times per shift, noted by supervisors as "efficient." The tea was the same brand available from the floor's automated dispenser. He selected it each morning without variation. Whether this was preference or the absence of the cognitive architecture required to prefer something else is not recorded.

Visual Identity

• Color palette: Brilliant red (#E74C3C) fading to institutional gray (#BDC3C7) — the thesis cover to the data entry floor
• Key symbol: A wireframe mind, half the connections severed — the remaining half still forming patterns nobody can read
• Lighting: Split — archive photographs show warm laboratory light; post-reduction images show flat fluorescence. The monitoring software logged both environments as "adequate illumination."