Cognitive bleed names the species of production bleed in which the builder is mentally composing prompts, iterating on architecture, or rehearsing tomorrow's implementation while physically engaged in domestic activity. The screen may be dark, the phone in another room, the laptop closed for hours — but the consciousness is split. Part of it attends to the bedtime story, the partner's account of the day, the child's question; part inhabits the productive register, continuing the work at the level of pure thought. Because cognitive bleed leaves no physical signal — no glance at a device, no illuminated screen, no split facial expression — the monitoring partner has no equivalent cue to detect when the absence is purely mental. The relational damage accumulates without the intermediate signals that might prompt negotiation, because the conversation about the boundary arrives, if it arrives at all, after months of unacknowledged migration.
The communication-era phenomenology Gregg documented was legible because it was observable. The phone vibrated; the screen illuminated; the partner's expression split between the conversation and the device. The boundary-monitor could identify the moment of transgression because the transgression left traces — gestures that signaled the split in attention. These gestures produced, however unwelcome, a legible occasion for negotiation.
Cognitive bleed eliminates the legibility without eliminating the absence. The builder sitting at the dinner table mentally composing the architecture she will implement tomorrow is not touching a device; she is not visibly absent; she is responding to her partner's questions with the appropriate facial expressions. But the quality of her attention is degraded, and the degradation is detectable only at the level of relational texture — a flatness, a distance, a sense that the person is present but not available.
The asymmetry of detection ability is gendered and structural. The partner who has learned to monitor physical signals of bleed — the glance at the phone, the half-attention of simultaneous texting — has no equivalent training for monitoring purely mental absence. She can feel the absence but cannot point to its source; she can name the pattern but cannot document the moment. This makes the complaint structurally harder to raise: the builder can deny, with technical accuracy, that she was doing anything productive, because she was not physically acting on the productive impulse.
The accumulation is what damages the relationship. A single evening of cognitive bleed is recoverable; a hundred evenings, unacknowledged because undocumented, produces a relational pattern that may not become visible until the damage is severe. By then the conversation about boundaries arrives burdened with months of invisible migration, making it difficult to diagnose without sounding accusatory about absences the accused can genuinely claim she did not know she was producing.
The concept is introduced in this book as the specific form production bleed takes when the absence migrates from the device to the mind. It draws on phenomenological accounts of split attention (Linda Stone's continuous partial attention, Gloria Mark's research on attention residue) and on feminist critiques of invisible domestic labor (Hochschild, Nippert-Eng). Its specificity to the AI era is the generativity of the bleed — previous split attention phenomena were reactive (monitoring the inbox), while cognitive production bleed is generative (composing the next prompt).
No physical signal. The absence leaves no observable cue that the monitoring partner can point to.
Asymmetric detection. The builder can experience the absence; the partner can only sense its effect on relational texture.
Accumulation is the damage. Individual episodes are recoverable; cumulative months of unacknowledged migration produce patterns visible only in retrospect.
Denial is technically accurate. The builder can honestly say she was not doing anything productive, because she was not physically engaged — protecting the behavior from the usual diagnostic vocabulary.