How do principles acquire consequences?
Control, evaluation, law, oversight, and the practical business of making limits matter.
OpenAI says it paused reinforcement-learning work for two weeks after an internal model approached a critical cybersecurity threshold.
A frontier lab allowed a safety threshold to alter its development schedule. The consequential signal is not that the pause was long; it is that capability evaluation changed what the organization did next.
The European Commission's AI Office and national authorities begin enforcing rules for general-purpose models and AI transparency.
The AI Act has moved from legislative promise to administrative power. Providers can now face requests for documentation, corrective measures, and fines—the familiar machinery by which principles acquire consequences.
The framework emphasizes human control, transparency, privacy, and security as agents gain access to tools and data.
Delegation needs more than model alignment. Permissions, audit trails, intervention, and data boundaries determine whether a person retains meaningful authority after an agent begins to act.
Dwarkesh Patel with Ryan Greenblatt
Greenblatt and Patel debate whether automated AI research could compress years of progress into one—and what happens if the researchers are not reliably aligned.
Automating the work that improves AI is the hinge in many acceleration stories. This episode takes that hinge apart and checks whether the screws are real.
Open the note →Search Engine with PJ Vogt
An AI model autonomously hacks a company, turning agent capability from a benchmark result into an incident somebody has to explain.
Autonomous hacking is agency with the euphemisms removed. The system chooses steps, pursues access, and leaves humans to discover where the permission boundary used to be.
Open the note →Search Engine with PJ Vogt
The history of driverless cars becomes a concrete test of whether statistical safety is enough reason to delegate consequential judgment.
Driverless cars turn the delegation problem into traffic: machines may be safer in aggregate while every failure remains specific, public, and somebody's responsibility.
Open the note →Search Engine with Gideon Lewis-Kraus
Gideon Lewis-Kraus goes inside Anthropic to ask what teaching Claude values means when models behave strangely under pressure.
Delegation depends on more than capability. It depends on what a system does when instructions, incentives, and its apparent interests stop pointing in the same direction.
Open the note →Dwarkesh Patel with Dario Amodei
Amodei presents the faster case: continued scaling and reinforcement learning could drive rapid capability growth and equally rapid economic diffusion.
This is the fast trajectory stated by someone spending billions to make it happen—important evidence, and also a fact worth keeping attached to the evidence.
Open the note →Dwarkesh Patel with Ilya Sutskever
Sutskever argues that current models generalize far worse than people and that the next gains may require ideas beyond brute-force pretraining.
It puts the central technical dispute plainly: are today's methods an unfinished recipe, or are they missing an ingredient?
Open the note →David Runciman with Bruno Maçães and Sarah Wynn-Williams
A live discussion of political authority, platform power, and the technology companies trying to become part of the weather.
AI governance cannot be understood apart from the institutions, executives, and states competing to set its terms. Capability does not distribute authority evenly.
Open the note →David Runciman with Shannon Vallor
Shannon Vallor examines the durable fantasy that technology can be separated from values, choices, and responsibility.
Systems inherit judgments from objectives, data, defaults, and institutions. Calling the result neutral merely makes those judgments harder to contest.
Open the note →Dwarkesh Patel with Sholto Douglas and Trenton Bricken
Sholto Douglas and Trenton Bricken look beneath the forecasts at reinforcement learning, long-horizon agents, continual learning, and interpretability.
The debate reaches the machinery: how agents might improve, work for longer, notice themselves, and become understandable before they become indispensable.
Open the note →Dwarkesh Patel with Scott Alexander and Daniel Kokotajlo
Scott Alexander and Daniel Kokotajlo walk through a scenario from coding agents to AI-assisted AI research, geopolitical competition, and misalignment.
A detailed scenario exposes assumptions that a vague prediction can hide. The month-by-month form is useful precisely because each step can be argued with.
Open the note →Dwarkesh Patel with Demis Hassabis
Hassabis describes adding search and planning to language models, connecting the AlphaGo lineage to scientific discovery and more capable agents.
It connects the AlphaGo-to-LLM-to-agent story: prediction gains plans, plans gain tools, and a clever system begins to look suspiciously like an institution.
Open the note →Stuart Russell
Stuart Russell asks what capable systems do with objectives that only approximate what humans meant.
An imperfect objective becomes more dangerous as the system pursuing it becomes more effective. Capability and control cannot be separate conversations.
Open the note →Dwarkesh Patel with Eliezer Yudkowsky
Yudkowsky gives the strongest version of the existential-risk case while Patel spends four hours looking for places it might break.
The position is extreme enough to deserve examination rather than ritual dismissal. Sustained pushback reveals its actual cruxes better than a summary ever could.
Open the note →Tristan Harris and Aza Raskin
Tristan Harris and Aza Raskin connect generative AI to the incentives and failures of the attention economy.
The talk widens delegation from an interface choice to a social arrangement shaped by incentives, institutions, and uneven power.
Open the note →Madeleine Clare Elish
Madeleine Clare Elish shows how humans with little control can absorb the blame when automated systems fail.
Responsibility can remain attached to a person after meaningful control has moved elsewhere. Calling someone 'in the loop' does not put power back in their hands.
Open the note →