<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Natasha Bajc on Medium]]></title>
        <description><![CDATA[Stories by Natasha Bajc on Medium]]></description>
        <link>https://medium.com/@natashabajc?source=rss-12a1fee76e4------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*A5Q7j_9cnDo6e-2M.jpg</url>
            <title>Stories by Natasha Bajc on Medium</title>
            <link>https://medium.com/@natashabajc?source=rss-12a1fee76e4------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 21:39:43 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@natashabajc/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Quantifying Durée: Vector Heat Mapping, Markov Chains, and Phenomenological Data]]></title>
            <link>https://medium.com/@natashabajc/quantifying-dur%C3%A9e-vector-heat-mapping-markov-chains-and-phenomenological-data-2a54f1cc5934?source=rss-12a1fee76e4------2</link>
            <guid isPermaLink="false">https://medium.com/p/2a54f1cc5934</guid>
            <category><![CDATA[architecture]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Natasha Bajc]]></dc:creator>
            <pubDate>Thu, 20 Aug 2026 16:28:22 GMT</pubDate>
            <atom:updated>2026-08-20T16:28:22.871Z</atom:updated>
            <content:encoded><![CDATA[<p>Bridging the gap between static architecture and lived duration (Kairos) using head-direction vectors and Markov chain spatial navigation models.</p><blockquote>“To bridge the gap between static architecture and lived experience, we must quantify durée without flattening it into Chronos.”</blockquote><p>Different concepts of time have existed since the beginning of time — pun intended.</p><p>The ancient Greeks recognized this by dividing temporal reality into two distinct dimensions: <strong><em>Chronos</em></strong>, the sequential, quantitative tick of the clock, and <strong><em>Kairos</em></strong>, the opportune, qualitative, and deeply felt moment of transformation. In our modern rush toward efficiency, we flattened time almost exclusively into <em>Chronos</em>. We treated time as a series of uniform, interchangeable seconds, forgetting that human consciousness inhabits <em>Kairos</em> and what philosopher Henri Bergson called <strong><em>durée</em></strong> (duration) — the unbroken, qualitative flow where past sensations continuously interpenetrate the present.</p><p>Architecture has long suffered under this flattened temporal paradigm: built space is static, but the experience of space is continuous, temporal, and fluid.</p><p>For centuries, architects attempted to capture <em>durée</em> through poetic intuition, subjective narratives, or post-occupancy surveys. But surveys are post-rationalizations; they record what an occupant <em>thinks</em> they experienced after the fact, filtering somatic reality through language.</p><p>To bridge the gap between static architecture and lived experience, we must do what previously seemed impossible: <strong>quantify <em>durée</em> without flattening it into <em>Chronos</em>.</strong> Through computer vision, head-direction vector extraction, and stochastic modeling, Architectural Artificial Intelligence (<em>A²I</em>) turns continuous phenomenological flow into actionable spatial telemetry.</p><h3>I. The Saccade vs. The Heading: Extracting Intentionality</h3><p>Most spatial analytics start and end with eye-tracking or simple foot-traffic heat maps. Both approaches fail to capture true somatic engagement:</p><ul><li><strong>Eye-Tracking (Biological Noise):</strong> The human eye produces 3 to 4 saccadic movements per second. Most of this is rapid, involuntary sampling — subconscious visual scanning that reflects physiological reflexes rather than deliberate spatial orientation.</li><li><strong>Foot-Traffic Heat Maps (Kinematic Footprints):</strong> Tracking footsteps records where a body <em>walks</em>, but reveals nothing about where its attention, orientation, or emotional interest is anchored.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*o80CltHo9kWk8wPx.jpeg" /></figure><p>In contrast, <strong>head-direction vectors</strong> operate at a motor-driven, intentional scale. Turning one’s head toward a threshold, an alcove, a skylight, or an artwork requires the recruitment of musculoskeletal effort. It is an act of <strong>motor intentionality</strong> — the physical alignment of the organism’s primary sensory apparatus with an architectural event.</p><p>By capturing egocentric head-direction vectors via non-invasive spatial vision models, we record genuine attention without burdening occupants with wearable apparatuses.</p><h3>II. Markov Chains &amp; Cognitive Landmarks: Uncovering Spatial Decision-Making</h3><p>Once directional vectors are extracted, how do we model the qualitative flow through space?</p><p>We structure the environment as a network of <strong>phenomenological anchors</strong> (spatial attractors, thresholds, compression zones, and vistas) and treat occupant navigation as a stochastic process using <strong>Markov chains</strong>.</p><p>A first-order Markov model calculates the transition probability <strong>P(S{t+1} | S{t})</strong> — the likelihood that an occupant oriented toward Anchor <em>A</em> will next transition their attention or posture toward Anchor <em>B</em>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*vk_nk-hGgI_IuzPk.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*AHH5fk8hZP5tI8Dq.jpeg" /></figure><p>Crucially, this modeling does more than record movement: <strong>it uncovers how humans conceptualize spatial navigation.</strong></p><ol><li><strong>Identifying Universal Internal Landmarks:</strong> Regions with intense vector convergence (high-density vector heat maps) reveal the features that the biological nervous system naturally prioritizes. These are not always what the architect intended as a centerpiece; they are the space’s emergent, universal <strong>cognitive landmarks</strong> — the anchors around which inhabitants mentally map the environment.</li><li><strong>Decoding Navigation Decision Trees:</strong> The transition probabilities between these landmarks expose the latent decision-making process of the human mind in space. When probabilities disperse evenly across multiple paths, the space induces cognitive friction and ambiguity; when they concentrate into clear transitional vectors, the architecture provides effortless intuitive orientation.</li><li><strong>Mapping Dwell States &amp; Friction:</strong> By computing the steady-state distribution of the Markov matrix, we discover where spatial attention naturally pools, recirculates, or prematurely drains.</li></ol><h3>III. Translating Egocentric Flow to Allocentric Models</h3><p>The ultimate destination for this phenomenological telemetry is the architect’s <strong>allocentric frame</strong> — the Building Information Model (BIM).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*-ycZVMYTHzmzKGtw.jpeg" /></figure><p>By streaming vector heat maps and transition matrices directly into BIM environments:</p><ul><li><strong>The Blueprint Becomes Dynamic:</strong> Static floor plans are overlaid with live or simulated vector density gradients, illustrating not just where bodies walk, but how they perceive boundaries and openings.</li><li><strong>Threshold Verification:</strong> Architects can evaluate whether a compression zone (like a low-ceiling entryway) successfully heightens the psychological impact of a subsequent decompression zone (like a double-height atrium).</li><li><strong>The Unthought Known Becomes Legible:</strong> Sub-perceptual friction — such as hesitation at an ambiguous split corridor or disorientation in an over-illuminated atrium — is pinpointed before costly post-occupancy retrofits.</li></ul><h3>IV. The Closed Loop of Experiential Design</h3><p>Quantifying <em>durée</em> is not about surveillance or Taylorist optimization; it is about <strong>epistemological accountability</strong>.</p><p>When we equip the architect with the tools to observe how lived duration and <em>Kairos</em> unfold within built form, design ceases to be an ungrounded guess. We replace the static snapshots of the past with a dynamic understanding of human embodiment and cognitive mapping.</p><p>By measuring the flow of bodily orientation and the stochastic paths of spatial navigation, we finally give architecture the empirical language it needs to design for the full spectrum of conscious life.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2a54f1cc5934" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Escaping the “Grey Goo”: AI as Lens, Not Creator]]></title>
            <link>https://medium.com/@natashabajc/escaping-the-grey-goo-ai-as-lens-not-creator-12870b188b38?source=rss-12a1fee76e4------2</link>
            <guid isPermaLink="false">https://medium.com/p/12870b188b38</guid>
            <dc:creator><![CDATA[Natasha Bajc]]></dc:creator>
            <pubDate>Mon, 17 Aug 2026 23:37:41 GMT</pubDate>
            <atom:updated>2026-08-17T23:37:41.606Z</atom:updated>
            <content:encoded><![CDATA[<p>Resisting generative homogenization through technosomatic phenomenology.</p><blockquote>“AI must not be the creator of ungrounded form, but a high-resolution phenomenological lens.”</blockquote><p>The modern design ecosystem faces an epistemological crisis: an unprecedented overproduction of frictionless “solutions” that are never evaluated. Generative AI tools and diffusion models churn out infinite spatial variations in seconds, sampling flattened 2D architectural precedents to generate buildings from buildings.</p><p>The result is a self-referential, aesthetically persuasive hallucination: <strong>the “grey goo” of disembodied form.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZZzUE0V94qdYfmvVMrveEg.jpeg" /></figure><p>The tragedy of this moment is not merely computational — it is institutional. Architecture as a discipline currently lacks an empirical method to evaluate its own hypotheses. When an industry has no mechanism to measure how a physical space modulates the human nervous system, every generative output carries the same weight. Infinite unverified forms collapse into statistical noise.</p><p>To escape this trap, we must redefine the purpose of machine intelligence: <strong>AI must not be the creator of ungrounded form, but a high-resolution phenomenological lens.</strong></p><h3>I. The Overproduction of the Untested</h3><p>Every architectural drawing is an implicit claim: <em>if we construct this threshold, this ceiling height, this material finish, human bodies will experience regulation, connection, or focus.</em></p><p>Yet spatial design remains one of the few high-stakes industries operating in an entirely open loop:</p><ul><li><strong>Aerospace and Software:</strong> Prototypes are subjected to continuous telemetry, automated testing, and iterative user feedback.</li><li><strong>Architecture:</strong> Buildings are designed, constructed, occupied, and demolished without ever returning structured somatic data to the architect’s desk.</li></ul><p>Generative AI entered this evaluation vacuum and weaponized it. By eliminating the friction of drawing, text-to-space and text-to-image models have supercharged production without introducing a single tool for assessment. We are mass-producing spatial proposals at machine scale while remaining entirely blind to their biological and cognitive impact.</p><h3>II. Academic Silos and the Resistance to Spatial Intelligence</h3><p>Why does architecture lack an evaluative mechanism? Because the discipline has guarded its borders against the very fields that could provide it.</p><p>Current institutional and academic politics have historically treated architecture as an insular discourse of formal composition, semiotics, or historicist theory. Cross-pollination from cognitive neuroscience, spatial biology, and phenomenological computing has been sidelined or viewed with skepticism.</p><p>By isolating itself from the science of how living organisms navigate physical environments, architectural academia has stalled the development of true <strong>spatial intelligence</strong> — an understanding of space grounded in embodiment, sensory processing, and neurobiology.</p><p>Without cross-disciplinary grounding, computational design is left with no standard of truth other than novelty. The academic engine continues to reward formal gymnastics while the real-world consequence — the human body moving through built space — remains untracked.</p><h3>III. The Inversion: AI as an Optical Instrument</h3><p>To break the cycle of the grey goo, we must invert the computational pipeline. We do not need AI to replace human intentionality in formal design; we need AI to give us vision where we are currently blind.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zExXe-hMznxDu9gVN4DCzg.jpeg" /></figure><p>Grounded in N. Katherine Hayles’s concept of the <strong>cognitive nonconscious</strong>, machine learning operates in the same sub-narrative, high-temporal register as our biological sensory systems. It can process patterns of human orientation that conscious reflection misses:</p><ol><li><strong>Extracting Intentional Orientation:</strong> Instead of generating speculative forms, computer vision tracks real-world <strong>head-direction vectors</strong> — the biological markers of motor intentionality.</li><li><strong>Quantifying <em>Durée</em>:</strong> By applying Markov chain models to physical movement transitions, AI translates Henri Bergson’s continuous <em>durée</em> (the lived flow of experience) into structured telemetry.</li><li><strong>Closing the Loop in BIM:</strong> Feeding this somatic data directly back into Building Information Modeling (BIM) tools converts architectural models from static geometry into dynamic, testable hypotheses.</li></ol><h3>IV. Beyond the Hallucination</h3><p>When AI is treated as a phenomenological lens rather than a generative vending machine, the role of the architect changes. The architect is no longer an editor sorting through algorithmic scrapbooks, but a systems orchestrator designing with full-board visibility.</p><p>The antidote to the grey goo is accountability to the human nervous system. By opening architecture to spatial neurobiology and using computational tools to rigorously evaluate spatial reality, we can stop manufacturing ungrounded novelties — and begin building spaces that anchor human life.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=12870b188b38" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Body as the First Screen: An Invitation to Technosomatic Architecture]]></title>
            <link>https://medium.com/@natashabajc/the-body-as-the-first-screen-an-invitation-to-technosomatic-architecture-e3606244d5fb?source=rss-12a1fee76e4------2</link>
            <guid isPermaLink="false">https://medium.com/p/e3606244d5fb</guid>
            <category><![CDATA[architecture]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[theory]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Natasha Bajc]]></dc:creator>
            <pubDate>Tue, 11 Aug 2026 15:24:40 GMT</pubDate>
            <atom:updated>2026-08-11T15:24:40.096Z</atom:updated>
            <content:encoded><![CDATA[<p>We have spent decades shrinking our cognitive landscape to fit inside glowing rectangles. In doing so, we treated human attention as a disembodied stream of text and visual tokens, forgetting that consciousness is anchored in a physical, spatial reality. We optimized for computational bandwidth while starving biological depth. We built networks that move at the speed of light, yet leave us feeling deeply ungrounded.</p><p>This platform exists to challenge that trajectory. It is an inquiry into <strong>Technosomatic Architecture</strong> — a theoretical and computational framework at the intersection of spatial neurobiology, phenomenological philosophy, and machine intelligence.</p><h3>Architecture as an Untested Hypothesis</h3><p>Architecture encodes a continuous hypothesis about the human body. Every section cut, ceiling height, and threshold is an implicit prediction: <em>if we build it this way, bodies will feel, move, and relate in this way</em>. Yet this hypothesis is almost never tested. Buildings are designed, occupied, and eventually demolished without returning somatic evidence to the architect. While software engineering and industrial design rely on iterative user feedback, spatial design operates in an open loop.</p><p>The recent generative AI turn has exacerbated this crisis. Current image synthesis tools and diffusion models generate spatial forms out of statistical noise — sampling millions of flattened precedent images to yield an aesthetically persuasive, geometrically hallucinated <strong>“grey goo”</strong>. They produce <em>buildings from buildings</em>, not <em>buildings from bodies</em>.</p><p>The fundamental flaw of modern computational design is a <strong>missing root node</strong>. In computer science, the root node is the origin from which all structural trees grow. Modern AI pipelines place prompt tokens or precedent images at the root. We must perform a radical correction: re-inserting the fully embodied human nervous system as the non-negotiable root node of every spatial data pipeline.</p><h3>The Technosomatic Imperative &amp; Spatial Neurobiology</h3><p>Human spatial intelligence is not a disembodied software algorithm; it is hardwired into our biology. Grounded in the Nobel Prize-winning research of May-Britt Moser and Edvard Moser, we know the brain navigates space through a complex internal positioning system — specifically place cells, entorhinal grid cells, and head-direction cells.</p><p>A vital historical and theoretical clarity must be maintained here: May-Britt Moser and Edvard Moser were co-researchers who jointly received the 2014 Nobel Prize in Physiology or Medicine. Honoring May-Britt Moser’s central leadership is key to a <strong>technosomatic, cyber-feminist stance</strong>: spatial intelligence is anchored in the somatic, lived reality of the body, resisting paternalistic over-simplification and disembodied abstractions.</p><p>While eye-tracking studies capture rapid, involuntary biological noise (3 to 4 saccades per second), <strong>head-direction vectors</strong> operate at an intentional, motor-driven temporal scale. When a person turns their head toward a spatial threshold, they are demonstrating motor intentionality — an active orientation of their entire attentional apparatus.</p><p>By utilizing AI as a high-resolution <strong>phenomenological lens</strong> (rather than a generative form-maker), we can extract and aggregate head-direction vectors from real-world spaces. Using Markov chain modeling to map transition probabilities between spatial anchors, we convert Henri Bergson’s continuous <em>durée</em> — the felt flow of lived experience — into precise, actionable design intelligence.</p><h3>The Parallax View &amp; The Fully Visible Board</h3><p>In <em>The Parallax View</em>, philosopher Slavoj Žižek describes irreducible gaps in perspective. In spatial design, a permanent parallax exists between two frames:</p><ul><li><strong>The Allocentric Frame (Architect):</strong> The bird’s-eye coordinate system of blueprints, plans, sections, and BIM models, which achieves clarity by abstracting away the human body.</li><li><strong>The Egocentric Frame (Inhabitant):</strong> The self-centered field of somatic affordances radiating from the individual body’s position, orientation, and heading.</li></ul><p>True spatial intelligence — what we call <strong>seeing the full board</strong> — is the ability to bridge this parallax. It translates egocentric somatic signals back into the allocentric spatial model, allowing the designer to navigate both positions simultaneously.</p><p>When you see the full board:</p><ul><li><strong>The “Unthought Known” is Made Legible:</strong> As psychoanalyst Christopher Bollas noted, the body possesses pre-linguistic knowledge of space — registering an architectural “glitch” or spatial discomfort long before conscious thought can name it.</li><li><strong>Embodiment Guides Systems:</strong> AI ceases to be a gimmick for rendering static geometry and becomes an instrument that measures how spatial configurations modulate human nervous systems.</li><li><strong>The Design Loop Closes:</strong> Architecture evolves from an untested, static artifact into an adaptive, somatically informed data ecosystem.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e3606244d5fb" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>