The Archive

User raises concerns about ID verification requirements and data privacy for Anthropic services.

Premium Claude subscriber complains of coding degradation in Opus 4.7 vs. 4.5, citing reliability issues despite Max plan subscription.

TSMC earnings data suggests skepticism within leadership about AI-driven growth forecasts and fab capacity expansion.

Berkeley BAIR proposes GRASP, a gradient-based planner for learned world models enabling longer-horizon planning via differentiable dynamics.

Tech workers in China are being instructed by their bosses to train AI agents to replace them—and it’s prompting a wave of soul-searching among otherwise enthusiastic early adopters. Earlier this month a GitHub project called Colleague Skill, which claimed workers could use it to “distill” their colleagues’ skills and personality traits and replicate them with…

Meme post contrasting lay perceptions of Copilot against technical audience anxiety regarding AI deployment.

User reports Claude assisted with home genome sequencing and wetlab setup without prior lab experience.

Discussion of arxiv paper volume growth and strategies for staying current with machine learning research.

NVIDIA shifts GPU kernel engineering tooling from C++ CUTLASS to Python CuTeDSL; debate on learning priorities for inference engineers.

SQL functions in Google Sheets to fetch data from Datasette instances using importdata(), named functions, or Google Apps Script.

User reports first prompt on Claude Pro exhausted daily usage limit and 13% weekly quota in 43-minute extended thinking session.

Claude Token Counter tool upgraded to compare tokenization across Claude Opus 4.7, 4.6, Sonnet 4.6, and Haiku 4.5.

Hyatt deploys ChatGPT Enterprise with GPT-5.4 across global workforce for productivity and guest experience improvements.

Analysis of headless API-first architecture replacing GUI interaction for personal AI agents, citing Salesforce Agentforce.

On the latest episode of Equity, we discuss OpenAI's latest acquisitions and whether they address "two big existential problems" for the company.

Vercel, a major development platform that hosts and deploys web apps, was compromised, and the hackers are attempting to sell stolen data. A person claiming to be a member of ShinyHunters, which was behind the recent hack of Rockstar Games, posted some data online, including employee names, email addresses, and activity time stamps. Vercel confirmed in a post on X that a "security incident" had occurred, and that it impacted a "limited subset" of its customers. Vercel said that a compromised third-party AI tool was the avenue for attack, though it did not specify which third party was involve...

A lot of AI startups exist partly because the foundation models haven't expanded into their category yet. As many jokingly acknowledge, that won't last forever.

Palantir's ideological bent has come under more scrutiny as it's worked with ICE and positioned itself as a defender of "the West."

Welcome back to TechCrunch Mobility, your hub for the future of transportation and now, more than ever, how AI is playing a part.

Curated list of ~1,200 ICLR 2026 papers (22% of 5,300+) with public code or data links.

Staff engineer considering career transition to research engineer role; seeks advice on feasibility.

Undergrad asks about future of spiking neural networks, neuromorphic computing, and liquid neural networks.

VIBE framework evaluates generative bias in Audio-Language Models using real-world speech and open-ended tasks instead of synthetic MCQs.

DORA Explorer improves LLM agent exploration diversity in sequential decision-making without training via entropy-based sampling.

CAMCO framework enforces policy constraints and auditability (SOX, HIPAA, GDPR) in multi-agent enterprise AI orchestration via constrained optimization.

HeadRank enables decoding-free LLM-based passage reranking by selecting attention heads with preference optimization to avoid score homogenization.

P-MLLM performs zero-shot personalized image aesthetics assessment using user profiles and selective fusion modules without historical rating data.

Project Yanasse discovers new mathematical proofs by transferring Lean 4 tactic patterns across Mathlib areas via GPU-accelerated analogy matching.

Empirical study comparing auxiliary loss interactions for training conditional depth routing gates in large language models.

MACE multi-agent framework verifies claims from tabular data using specialized Planner, Executor, Verifier agents with zero-shot CoT.