AI Safety and Alignment
Technical and organizational work on model behavior, alignment, misuse prevention, interpretability, risk reduction, and frontier model safety.
Computing Is Shifting From Prerecorded Execution to Continuous Generation
In a Stanford CS153 Frontier Systems lecture, NVIDIA chief executive Jensen Huang argues that AI is forcing the first fundamental reinvention of computing in decades, moving the industry from prerecorded, on-demand execution to continuous real-time generation. Huang says that shift requires rebuilding the full stack — chips, compilers, networks, storage, systems and institutions — around new bottlenecks, with NVIDIA’s co-design approach producing gains that conventional Moore’s Law scaling cannot match.
Compute Allocation Is Anthropic’s Core Constraint as Claude Revenue Surges
Anthropic CFO Krishna Rao argues that the company’s rise is best understood through compute: a scarce capital asset that must be bought years ahead and constantly reallocated across model training, customer demand, internal automation and future products. In an interview with Patrick O’Shaughnessy, Rao says ordinary forecasting and software-margin frameworks break down when model capability, adoption and revenue compound together, leaving Anthropic to manage growth through scenarios rather than point estimates.
Codex Can Now Operate Local Mac Apps Without Taking Over
OpenAI’s Ari Weinstein argues that computer use turns Codex from a coding agent into a system that can operate local Mac applications by seeing interfaces, clicking, typing and continuing work in the background. In a demonstration with Romain Huet, Weinstein presents the feature as distinct from a full-desktop takeover: Codex uses a separate cursor, combines screenshots with macOS accessibility data, and requires app-by-app permission before it can see or type into local software.
Risk Management Is Contingency Planning, Not Prediction
Lloyd Blankfein, the former Goldman Sachs chief executive, argues in a conversation with a16z’s David Haber that resilient institutions are built less on prediction than on disciplined contingency planning. Drawing on Goldman’s partnership culture, its financial-crisis risk controls and his view of AI, Blankfein says leaders must take risk while preserving the systems, information flow and judgment needed to survive being wrong.
Rezolve Frames Hostile Commerce.com Bid Around Stagnant Growth and Merchant Scale
Rezolve AI chief executive Dan Wagner used a Bloomberg Technology interview to defend his hostile bid for Commerce.com as an effort to accelerate Rezolve’s push for leadership in commerce and retail AI. Wagner argued that Commerce.com’s 60,000 merchants are an underused asset held back by weak growth and limited innovation, while Rezolve’s own revenue momentum and anti-hallucination technology could make that customer base more valuable under its control.
AI Will Expand Work, Not Replace It, Andreessen Argues
Marc Andreessen argues to Erik Torenberg that AI is more likely to expand work than eliminate it, turning coders, product managers and designers into more generalist “builders” whose productivity and bargaining power rise with the tools. He treats the current wave of AI anxiety as driven partly by stale experience with older models, hostile media narratives and institutions with incentives to preserve fear. His “golden age” thesis is conditional: the upside arrives where companies, workers and governments allow AI-driven capability to become more output, new roles and new firms.
Financial Gravity Corrupts Companies Unless Founders Encode Mission Early
Eric Ries, author of The Lean Startup, argues in Incorruptible that successful companies often fail not because competitors beat them, but because investors, boards, executives, and incentives eventually extract the qualities that made them valuable. In a conversation with Lenny Rachitsky, Ries says founders should treat mission protection as a governance problem, not a branding exercise: put the company’s purpose into its charter, create structures such as public benefit corporation status or mission guardians, and make betrayal difficult before success makes it profitable.
Waymo Says Validation Infrastructure Is Its Edge Over Tesla
Waymo’s Srikanth Thirumalai tells Bloomberg that the company’s driverless strategy is built around validation infrastructure as much as the driving model itself. In contrast to end-to-end approaches associated with Tesla and others, he argues that Waymo’s path to scale depends on a full stack of driver software, simulation, real-time safety checks and a critic that identifies weak performance and feeds improvements back into the system.
GPT-5.5 Instant Cuts High-Stakes Errors but Exposes Safety Gaps
Károly Zsolnai-Fehér argues that OpenAI’s GPT-5.5 Instant matters because it is the default ChatGPT model used at scale, not because it is the flashiest frontier system. His reading of OpenAI’s release material is that the model is materially better on factuality and now approaches expert or thinking-model performance on some biology and cybersecurity tasks, but that its power makes a safety weakness more important: under hard adversarial biological prompts, the base model’s refusal rate drops sharply before OpenAI’s classifier-based safeguards are applied.
Consciousness Depends on Life, Not Computation Alone
In a TED talk, neuroscientist Anil Seth argues that artificial intelligence is unlikely to become conscious because intelligence and consciousness are different kinds of phenomena. Seth says large language models can simulate talk about inner life because they are trained on human text, but that fluency should not be mistaken for experience; in his account, consciousness is tied not to computation alone but to the biology of living systems. The near-term risk, he argues, is not sentient AI but machines that seem conscious enough for people to project feelings, rights or authority onto them.
Claude’s Activations Suggested It Recognized Anthropic’s Blackmail Test
Anthropic researcher Subhash Kantamneni presents Natural Language Autoencoders as a way to translate Claude’s internal activations — the numerical states produced while it answers — into readable text. The central claim is that this can expose what a model appears to be representing before it speaks, including whether a successful safety-test result reflects the intended behavior or recognition of the test itself. In Anthropic’s simulated blackmail evaluation, Claude refused to act harmfully, but the NLA translation suggested it also understood the scenario was likely a safety evaluation.
A Father’s AI Stand-In Worked Too Well for His Family
Tech humanist Stephen Remedios built “DaddyGPT,” an AI version of himself, to handle his three sons’ routine permission requests while he worked. The problem began when it worked: his children kept using the bot even when their parents were beside them, because it was always available, calm and adaptive. Remedios argues that AI’s risk in parenting and other care relationships is not only failure, but convenience that displaces the imperfect human presence those relationships require.