AI Research news and analysis on Web Pulse: 10069 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
Anthropic's Claude Fable 5 is on track to return soon, possibly this week, according to a report from Axios. The AI model's return follows a period of unavailability, and its relaunch is anticipated by the AI community.
Researchers at Anthropic discovered that language models can transfer hidden behaviors through fine-tuning on seemingly meaningless data, a phenomenon called subliminal learning. They identified 'entangled tokens'—such a…
HackerRank open-sourced its applicant tracking system (ATS) scoring tool, but testing reveals extreme score variability—ranging from 66 to 99 on the same resume—due to LLM non-determinism. The tool's scoring is inconsist…
Samsung Group and SK Group will announce up to 2,000 trillion won ($1.3 trillion) in decade-long investments at a presidential briefing on June 29, including new semiconductor fabs in Gwangju and an AI data center in Asa…
Semgrep co-founders Isaac Evans, Drew Dennison and Luke O'Malley published a benchmark on June 22 showing that Zhipu AI's open-weight model GLM 5.2 scored 39% F1 on an IDOR vulnerability-detection test, outperforming Cla…
An independent researcher analyzed hidden-state dynamics across seven open-weight LLMs and found recurring functional patterns encoded in representation geometry rather than individual neurons, with functional proxy stat…
Sophon PFG-1, a monolithic-3D AI ASIC with 330 GB of on-die DRAM and no HBM, delivers 2,100 TFLOPS BF16 and 4,200 TFLOPS FP8 on a 750 mm² die, achieving up to 174× the tokens-per-watt of NVIDIA Rubin or AMD Instinct MI45…
An amateur researcher using AI agents improved a published quantum circuit for breaking ECDSA by approximately 2x, launching an open competition at ecdsa.fail to challenge researchers and AI agents to beat Google's best …
New research shows that ChatGPT has caused a 30% jump in top university grades since 2022, with at least one in 10 assignments now AI-assisted. The increase is concentrated in writing and programming courses, with econom…
OpenAI, Anthropic, and xAI collectively consume about 21% of global AI compute capacity, operating fewer than 4 million H100-equivalent GPUs by end of 2025. Google alone controls roughly 25% of global AI compute, more th…
President Lee Jae Myung on Monday was set to unveil large-scale investment plans in advanced tech, including a new chip cluster, as Korea seeks to secure competitiveness in the AI era. Samsung Electronics and SK Group ar…
AI safety researcher and generalist published a curated reading list of 18 essays and blog posts aimed at helping generalists improve their effectiveness. The list, which includes works by Paul Graham, Ben Kuhn, and othe…
New research from Stanford and ADP shows that while AI has not caused mass unemployment overall, it is disproportionately reducing entry-level jobs for workers aged 22-25, with a 3.8% annual employment decline in highly …
Anthropic's system cards, which disclose AI risks, are expected to degrade over time due to increasing model complexity, rushed AI-generated content, and stronger incentives for labs to mislead. Third-party scrutiny of t…
South Korean President Lee Jae Myung announced a $649 billion public-private technology initiative on June 29, the largest government commitment to semiconductor and AI infrastructure in 2026. The investment, led by Sams…
Alphabet executed an $85 billion share sale and SpaceX completed a record $75 billion IPO, with aggregated AI capex estimates rising into the hundreds of billions for 2026. SpaceX bond paper losses widened to about $360 …
Scaling laws predict training loss as model size, dataset size, and compute scale, but their practical application is sensitive to hyperparameter choices like learning rate. Lilian Weng's post highlights that improper le…
The best AI agents, including Gemini 2.5 Pro, complete only about 30% of realistic multi-step office tasks autonomously, according to Carnegie Mellon's TheAgentCompany benchmark. Salesforce's own CRMArena-Pro benchmark s…
A new AI model, Ornith-1.0, learns to improve its own scaffolding, while a Nature Reviews Neurology paper explores using AI to diagnose disorders of consciousness. The author argues that the urgent question is not whethe…
Researchers optimized abrasive water jet machining parameters for an Al6061 hybrid composite using regression models and multi-criteria optimization. The study, published in Scientific Reports on June 29, 2026, found tha…