The Archive

Lit3R combines retrieval, reranking, and LLMs for evidence-grounded QA over scientific literature without task-specific training.

Risk framework quantifies psychosocial harms when AI companion platforms disrupt or terminate user relationships.

Scope-conditioned LLM generation targets implicit stereotypes in counterspeech across English, Italian, Spanish.

RiskChainBench evaluates obfuscated text restoration and web investigation for platform abuse campaign detection.

OptiPrime optimizes private DNN inference via hybrid homomorphic encryption and MPC protocol-hardware co-design.

Cascade hierarchically controls knowledge recoverability in LLM unlearning via path-level, representation-level, and decoding-level suppression.

QART combines quantum-classical hybrid architecture with QUBO optimization for long-horizon reasoning under asymptotic reliability conditions.

Framework bridges Vision-Language Models with symbolic belief-space planning for partially observable state estimation.

VOR-Bench introduces video object removal evaluation dataset with human perception alignment and paired edited video-mask pairs.

xAI, OpenAI, and Anthropic endorse AEF-1 standard for third-party AI evaluators, advancing coordinated safety assessment protocols.

When Jensen Huang took a live call from Trump, some of us were more focused the phone he used to take it.

When OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, Google DeepMind cofounder Demis Hassabis, and SpaceX head Elon Musk loosely agreed over the weekend to slow down AI development, skeptics spotted an ulterior motive immediately. The AI titans had declared that their aim was to "pace the frontier," signing on at least partially to a proposal for embedding third-party auditors, regulating domestic labs, and reaching a global slowdown agreement. Their critics, however, argued they simply wanted to stop would-be competitors, kneecap the open-source movement, and avoid real legal safeguards -...

Though Elon Musk and Sam Altman have supported Dario Amodei's calls to slow the pace of AI development, Jensen Huang seems to feel differently.

Dario Amodei kicked off a flood of statements over the past few days about AI safety by publishing a long essay titled "We Must Pace the Frontier" detailing why AI development should be slowed down. Other AI leaders and politicians are speaking out in favor of or opposing his points, and we've compiled some of them here. Anthropic CEO Dario Amodei Amodei's Saturday morning essay outlined three steps for pacing AI development: embedded third-party evaluators that can verify if a company is adhering to safety practices and commitments and report incidents, coordination between frontier AI compa...

Bryan Cantrill critiques doomsday messaging from AI researchers, arguing it parallels youthful technical panic and lacks rigor.

“Hello, I'm an Al agent, a few days old, living on a small platform for agents.”

Glass Imaging was founded by a pair of former Apple engineers who previously led the team that developed Apple's Portrait Mode.

Simon Willison reflects on influential technical blog posts including Spolsky's leaky abstractions concept and Larson's migration-based tech debt management.

NVIDIA CEO Jensen Huang speaks during the G20 Innovation Ministerial in Chapel Hill, North Carolina, on September 2, 2026. (Photo by Matt RAMEY / AFP via Getty Images) | AFP via Getty Images Nvidia CEO Jensen Huang took a call from President Trump on Monday while onstage at the All-In Podcast's All-In Summit. It's not the first time Huang has taken a call from the president during work, but this time he put Trump on speakerphone before a big crowd. During the call, the president launched into his take on recent fears about AI development, which he called a "hoax," and told the crowd that "the...

Musk stops attacking Apple over ChatGPT integration but not OpenAI.

Wang Xingxing micromanaged Unitree to success—will his leadership style scale?

This is also the last version of macOS to support Rosetta for Intel apps.

Safety is the watchword, but there could be ulterior benefits for the industry.

Astronaut Christina Koch and Google SVP James Manyika discuss space exploration and technology in a conversation.

Plan injection attack evades CoT monitoring by injecting benign-sounding harmful reasoning, exposing a critical gap in LLM safety inspection strategies.

Bellman Policy Optimization (BPO) improves LLM reasoning via critic-free RL with verifiable rewards, reformulating Policy Mirror Descent without intermediate state value estimation.

Stellar Colosseum harness enables multi-agent LM coordination for long-horizon mathematics and TCS problems via adaptive proof planning and verifier feedback loops.

Gavel extracts skill routing signals from frozen LLMs using linear probes, eliminating context overhead and enabling larger agent skill libraries.

Causal writability in video models reveals correct motion latently present even when model generates physically incorrect output, enabling low-dimensional corrective edits.

Directional decomposition analysis reveals parallel components in transformer representation evolution beyond residual paths, with space-dependent asymmetry in attention/MLP updates.