The Austin Convention Center is rubble. The sessions are scattered across a hotel-grid not built for this load. Signal still came through. Two panels on cognition and one on space infrastructure kept circling the same question from different altitudes: what survives the merge?
Opening Pulse
The Austin Convention Center is gone. Demolished. The replacement isn’t finished yet, so SXSW 2026 is running on a distributed grid of hotel ballrooms, overflow lines, app streams, and human rerouting.
Day one felt overloaded immediately.
I caught the first two sessions through the app. Phone pressed against my face and recording. Standing against walls. Moving between buildings while crowds pressed in from every direction.
You find a method. You adapt the interface. The sessions happen. So do the notes.
// sys.environment
convention_center = null
venue_model = distributed
throughput = unstable
adaptation requiredThe physical infrastructure is degraded. Attention infrastructure matters more because of it.
Sessions
AI & the Brain: As We Embrace AI, Let’s Not Forget Our Minds
// session.log
Olivia Joseph // MIT // Computation + Cognition
Sanjay Sarma // MIT // Mechanical Engineering
Chris Gabrieli // Massachusetts Board of Higher Education
Izzat Jarudi // EdificiiThis panel named something a lot of people in AI rooms seem reluctant to say out loud: the writing is flattening.
Olivia Joseph put it plainly. LLM-shaped language is converging toward a shared cadence, a shared register, a shared confidence pattern. Homogenization at its finest.
Sanjay Sarma widened the frame. Education is drifting into a bad feedback loop where assessment drives curriculum, even when the assessments no longer map cleanly to what matters. Chris Gabrieli pushed the institutional layer harder. Higher education is operating at reduced capacity, while public faith in the degree-as-credential model continues to erode. The system is straining in multiple places at once.
Izzat Jarudi brought Kahneman into the room. Fast thinking is exactly where AI excels. Slow reasoning still belongs to the human, at least for now. Gabrieli ended on the most ambitious note of the session: a call for a Framingham-style longitudinal study on what AI use is actually doing to cognition over time.
That study does not exist yet. It probably should.
Key points
Language homogenization is becoming measurable, not anecdotal
Higher ed is strained structurally, financially, and culturally
Skills-based hiring remains more slogan than operating reality
Competency-based education continues to surface as a possible counter-model
AI literacy is baseline now, not advanced
A long-horizon cognitive study on AI use is missing and urgently needed
// Cyborg note: The flattening of voice is a symptom. The deeper problem is a cognition loop that stops producing signal worth processing. When output converges, the next training cycle gets thinner too.
The Learning Geeks: How to Think When Machines Think Faster
// session.log
Bob Gerard // Accenture
Jake Gittleson // McKinsey Learning Lab
Dr. Tessa Forsythe Shaw // Harvard GSE / Next Level LabThis one hit harder.
Tessa Forsythe Shaw brought the strongest frame of the day: the brain is metabolically expensive, and humans are wired to offload effort whenever possible. Not laziness but organism logic.
Generative AI changes the environment around that instinct. It removes many of the triggers that force metacognition online. Struggle. Partial understanding. Confusion. Unfinished thinking. The tool returns something coherent and complete, and the loop that would have asked the human to wrestle with the problem never fully engages.
She returned several times to the idea of productive friction. Not difficulty for its own sake. The specific kind of friction that produces learning. The messy middle. The unresolved layer. The place where reflection happens during the work, not only after the work is done.
Several studies surfaced here, and they all rhymed. Higher trust in AI tends to correlate with lower critical engagement. Junior developers using AI understand less of their own code. AI-assisted professionals can begin to deskill quickly when cognitive strain gets removed from the system.
This session had the sharpest design implication of the day.
The interface is the intervention point.
Key points
Metacognition remains a major predictor of learning and creative performance
“Cognitive surrender” is showing up as a real pattern, not a cultural vibe
AI can remove the environmental cues that trigger reflective thinking
“Elicit before generate” feels like an important design principle
Calibration matters: users need some sense of their actual competence to benefit from AI without dissolving into dependence
// Cyborg note: A tool that preserves human cognition while assisting it is a design problem. Still mostly unsolved. Very worth solving.
Exploring the Future of Space & Japan’s Deep-Tech
// session.log
Will Rice // Cesium Astro
Jared Corrigan // Woven Capital
Tomoki Osato // Sora Materials
Nanak Hiram // Deloitte TohmatsuThe altitude changed. The question didn’t.
This panel framed space as infrastructure, not experiment.
Cesium Astro is building modular, software-defined communications payloads for space systems. The old bespoke aerospace model does not scale cleanly into proliferated low-earth orbit. Woven Capital framed the opportunity through mobility. Information, people, goods, and energy all depend on connectivity, and terrestrial infrastructure alone cannot carry the future load cheaply enough.
Tomoki Osato brought the room’s material imagination online. A composite that floats. Thermal insulation. Electromagnetic shielding. Sound absorption. A startup material born from university flight experimentation and now pointed toward space-capable applications.
That was the jaw-drop moment, but the deeper signal sat elsewhere.
Austin’s RF depth, semiconductor ecosystem, software culture, and proximity to defense networks are no longer adjacent facts. They are fusing into a regional systems advantage. Meanwhile, Japan’s deep-tech ecosystem is scaling through manufacturing strength, startup infrastructure, and long-horizon national coordination.
The design conversations surrounding this world still feel behind it.
Key points
Space is being framed as a practical infrastructure layer, not a frontier fantasy
Modular, reconfigurable systems are replacing bespoke mission-by-mission builds
Cross-domain mobility depends on non-terrestrial connectivity more than many UX conversations currently acknowledge
Material innovation is becoming a systems story, not just an engineering one
The defense, venture, and infrastructure layers are converging faster than design discourse has caught up to
// Cyborg note: Sovereign capability and startup culture are merging at the infrastructure layer. The design instincts that survive that merger will shape the next twenty years of built environment.
Day Theme
Two sessions on cognition. One on space infrastructure. Same question, different altitude.
When the machine gets better, what part of the human gets preserved?
The morning sessions suggested the answer is unstable without intention. The afternoon session suggested at least some of the people building the next hardware layer are still thinking about what it is for.
That question matters more than whatever the AI cliché of the week happens to be. The cliché is already inside the system.
// sys.daylog
infrastructure = degraded
capacity = exceeded
signal = partial but stable
attention rerouted successfullyClosing
Tomorrow’s targets include two futurists, entertainment IP, and the so-called space superhighway.
That is the plan, anyway.
Day one was a reminder that plans at SXSW are provisional. Overcapacity changes the route. Rooms fill early. Lines become interfaces of their own. Some of tomorrow will come down to game-time decisions and available bandwidth. Maybe that settles after opening-day surge. Maybe it doesn’t.
Either way, today’s working question emerges:
If metacognition is part of the antidote, what does a tool designed to preserve it actually look like?