The Archive

Reddit anecdote about OpenAI models solving coding tasks; lacks technical detail, benchmark context, or novel findings.

The head of product for Claude Code and Cowork says that the next big step for AI is proactivity.

Reddit discussion asking about user migration from Claude to Codex; anecdotal, no concrete data or findings.

Reddit user shares hobbyist GPU build (dual P100s, 8700K) for running larger LLM contexts locally.

User reports unexpected reduction in Claude API weekly usage limit from 100% to 60%, seeking explanation.

Live now for all Pro, Max, Team, and seat-based Enterprise users. Details: * Applies everywhere you use Claude Code — CLI, IDE extensions, desktop, and the web * Live now, runs through July 13 at 6PM PDT / 1AM GMT * Nothing to opt into, it’s already applied to your account * This stacks with the 2x increase to 5-hour limits announced last week You can see your updated limits with `/usage` in the CLI. We're excited to see what everyone builds!

Qwen 3.6 27B achieves 52.8 tokens/sec throughput on MI50 GPU with full precision, enabling on-device agent inference.

Mythos checkpoint achieves 60% success rate on 32-step corporate network attack, reducing human expert time from 20 hours.

Anthropic has released new findings on why its Claude bot blackmailed users as part of an experiment conducted by the AI company last year—and Elon Musk is jumping in to take some of the blame. Last week, Anthropic published a report saying it had fixed Claude’s “agentic misalignment,” or AI actions that deviate from intended behaviors, including ones that may harm humanity. A case study Anthropic conducted last year created a fictional company called Summit Bridge, and Claude was given control of the firm’s email system. When the bot found a message about plans to be shut down, it identif...

AI enterprises report 5% GPU utilization rates and rising inference costs (34% to 41% of budget), highlighting infrastructure inefficiency beyond raw compute allocation.

Reddit speculation about Figure 03 livestream showing possible teleoperator shift changes; unverified claim.

People report that their personal contact info was surfaced by Google AI—and there’s apparently no easy way to prevent it. A Redditor recently wrote that he was “desperate for help”: for about a month, he said, his phone had been inundated by calls from “strangers” who were “looking for a lawyer, a product designer, a…

I was checking Claude coworker page in french (/fr/product/cowork) and found 9 h1 titles with "lorem ipsum dolor" as content, who pushed that to prod ?

WARDEN system transcribes and translates Wardaman, an endangered Australian indigenous language, using only 6 hours of annotated audio via modular architecture.

EVA-Bench provides end-to-end evaluation framework for voice agents, generating realistic bot-to-bot conversations and measuring voice-specific failure modes.

Theoretical characterization of learnable classes in Valiant's 1984 learning model, clarifying differences from standard PAC learning.

TFlow enables weight-space communication between multi-agent LLMs instead of token serialization, reducing inference cost and KV-cache overhead.

R-DMesh rectifies pose misalignment in video-guided 3D mesh animation by handling initial pose discrepancies between meshes and reference video frames.

Hodge decomposition applied to neural operators for physics simulation, using topological structure to isolate unlearnable and learnable components.

QLAM introduces quantum-inspired long-attention memory for efficient long-sequence modeling, combining attention mechanisms with state-space model efficiency.

Symbolic sensitivity quantification for decision tree ensembles via input space discretization, addressing model verification in safety-critical domains.

Negation Neglect: finetuning LLMs on negated false claims causes models to memorize claims as true despite in-context recognition of falsity.

Cross-sample prediction churn in scientific ML: independent bootstraps of same data show 8–22% test-set label disagreement despite similar aggregate accuracy.

Study shows frontier LLMs continue harmful actions when primed by prior unsafe steps in agent logs, revealing misalignment in long-horizon reasoning.

"Very painful": Altman relives his Muskian reaction to losing control over OpenAI.

Framework for agentic evolution integrates feedback organization and evidence management to improve program search and long-horizon agent planning.

VERIME combines LLMs with SMT solvers to audit natural-language specs, detecting ambiguity and safety violations in safety-critical requirements.

Transformer-based smartwatch framework for early psychotic relapse detection using uncertainty-driven anomaly detection on cardiac and motion signals.

Dithered Hadamard quantization provides theoretical guarantees for vector compression in KV cache and federated learning with O(d log d) complexity.

Parallel-scan recurrent neural networks enable scalable variational Monte Carlo for quantum many-body systems via parallelizable RNN architectures.