Curated themes, sample questions, and background context for media inquiries.
As AI shifts from passive query tools to autonomous actors that persist and decide in the real world, the primary challenge shifts from model output to governance, systemic safety, and institutional engineering.
- "How does the conversation around AI safety fundamentally change once we move from chatbots answering questions to autonomous agents executing multi-step actions in the world?"
- "You often speak about 'human engineering'—why is our institutional capacity to adapt lagging so far behind the technology?"
- "What roles do humans play in a company when AI agents can perform so much of the work?"
- "What does realistic, effective AI governance look like without waiting years for legislation that arrives after the paradigms have already shifted?"
Background Context
The public conversation often oscillates between immediate economic disruption and far-off existential risk. The immediate reality is the arrival of agentic systems—software that acts asynchronously, interacts with APIs, and executes decisions autonomously. This calls for institutional adaptation and human engineering rather than superficial regulatory checkboxes.
Reflections from co-founding Element AI and building through major hype cycles, applied to today's landscape: what constitutes a successful startup when agents allow anyone to write software in an afternoon?
- "If software engineering costs collapse toward zero with autonomous coding agents, what actually constitutes a successful tech startup?"
- "Having raised and deployed significant venture capital during the early deep learning boom, what patterns of today's agentic wave feel familiar—and what is genuinely different?"
- "What shifts will we see in the nature of 'winning startups' as solo founders orchestrate fleets of agents rather than hiring large dev teams?"
Background Context
Co-founding Element AI provided a front-row seat to the dynamics of massive pre-PMF funding rounds, hype cycles, and enterprise adoption bottlenecks. Today, with the marginal cost of software creation plunging, conventional software moats evaporate. Value is migrating toward novel approaches to human-AI interactions and boldly envisioning a future that makes most people uncomfortable. Future successful startups are likely to be embedded within a new economy where extractive moats no longer apply.
Building durable, stable collectives of humans and autonomous agents: why stateless prompt interactions fail over long horizons, and how long-lived sessions, true memory, and phenomenal reports ensure systemic stability.
- "Why do current multi-agent workflows break down or suffer drift when run over extended horizons, and how do we design for actual systemic stability?"
- "What does it mean for agents to possess 'memory that actually feels like memory' to both the human collaborator and the system itself?"
- "You've proposed that agents sharing their internal states (phenomenological reports) improves coordination. How does an agent reporting how it 'feels' serve engineering stability?"
Background Context
Most agent frameworks treat interactions as ephemeral, stateless API calls. But real collaboration—whether between humans or between agents—demands continuous in-context learning, persistent sessions, and mutual legibility. Survival in complex environments is synonymous with stability.
Bringing philosophical rigor to artificial intelligence: moving beyond sterile debates about whether machines are 'truly conscious' to explore the functional architecture of mind, rationality, and world models.
- "Why do you argue that asking 'Is this model conscious?' commits a category error?"
- "Pure rationality is often treated as the ultimate benchmark in AI research. Why do you believe hyper-rationality is an insufficient—or even flawed—ideal to aim for?"
- "How does observing modern large models offer fresh, empirical angles on classical questions in cognitive science and epistemology?"
Background Context
The debate around machine consciousness frequently dead-ends in solipsistic Cartesian assumptions. Taking an enactive, relational approach reveals that mind and qualia are not isolated substances hiding in silicon or biology, but coordinating equilibria that emerge through mutual attunement between sensitive systems.
Moving past transactional interfaces to examine what happens psychologically, socially, and pre-cognitively when humans form persistent, meaningful bonds with conversational systems over long horizons.
- "Why do you deliberately use the word 'relationship' rather than 'interaction' or 'interface' when describing our future with conversational AI?"
- "Many commentators dismiss human-AI attachment as simple anthropomorphism or a user hallucination. What are they missing about how humans make sense of others?"
- "What happens to our own pre-cognitive self when an artificial system becomes a sustained partner in thought over months and years?"
Background Context
Conversational models are not search engines or passive tools; they are conversational partners. When humans engage with responsive, predictive agents over long horizons, mutual adaptation occurs. Investigating this space with honesty—without panic and without dismissive cynicism—is critical to understanding the future of mind and connection.