AI Research news and analysis on Web Pulse: 8407 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
Indonesia's National Research and Innovation Agency (BRIN) and the Ministry of Higher Education, Science, and Technology are drafting a national research roadmap for 2026–2045, set to launch in August. The roadmap will p…
OpenAI CEO Sam Altman has long advocated for humanity to 'merge' with artificial superintelligence, suggesting brain uploading as a path to survival. Critics argue this digital afterlife would only be accessible to the u…
Researchers at Goodfire, Harvard, Stanford, and Brown introduced Block-Sparse Featurizers (BSF), a method that decomposes neural network activations into multidimensional subspaces instead of single directions. Applied t…
Samsung has begun mass production of PCIe 6.0 SSDs with up to 64TB capacity for Nvidia's Vera Rubin AI platform, which entered full-scale production on June 1, 2026. The drives, including the PM1763 and PM1753 variants, …
Anthropic revealed that its AI model Claude can silently perform reasoning steps in a conceptual space called J-space, enabling it to detect code bugs and identify images. The discovery highlights a new capability in AI …
Meta's Muse Image model rose to second place on Arena.ai's Text-to-Image Arena leaderboard with an Elo score of 1280, trailing only OpenAI's GPT Image 2 at 1385. The model, announced by Meta Superintelligence Labs, featu…
A developer argues that AI evaluations are structurally identical to traditional software tests, just rebranded with new terminology. The post claims that the eval-tooling stack—golden datasets, judges, scorers, rubrics—…
A new analysis by Lily Ray reveals that self-promoting listicles in AI search often backfire, with Google's AI Overview citing a brand's own page but recommending a competitor 69% of the time. The research shows that rec…
Anthropic released Claude Opus 4.7 on April 16, 2026, a targeted upgrade focused on making agentic coding loops and multi-step tool use production-ready. The model achieves 87.6% on SWE-bench Verified and 64.3% on SWE-be…
Z.ai's open-weight GLM 5.2 completed an agentic coding task in 17 minutes at $2.76, while Anthropic's Claude Fable 5 finished in 9 minutes at over $10, with comparable output quality. The open model cost roughly a quarte…
A new study from the barkup-bench series reveals that LLM agents suffer from session drift when serializing tree state only once, with per-step success dropping from 98.8% to 83.8% for Claude Sonnet 4.5 over 12 turns. Th…
Yggdrasil Agent, a general-purpose parallel reasoning system, launches with a multi-agent architecture that splits work across ten cognitive realms to improve accuracy and reduce blind spots. The system is vendor-agnosti…
Researchers have discovered a region within large language models called 'J-space' where conscious-like reasoning occurs, using a new interpretability technique called the Jacobian Lens. The J-space enables verbalizable …
Chinese AI models are gaining traction among U.S. companies as costs for American alternatives like OpenAI and Anthropic surge, according to a CNBC report. The trend highlights growing adoption of Chinese AI technology i…
Canada's trade surplus widened to $4.2 billion in May, driven by higher sulphur shipments. The CRTC ordered Bell, Rogers and Telus to justify new fees by July 30. Scotiabank, Sun Life, Telus and Lightworks launched an AI…
A lonely girl named Alice, home for the summer, dreams of having a twin to share her imaginative adventures and secret societies, reflecting her desire for a companion who truly understands her.
Tech CEOs and billionaires are descending on Sun Valley, Idaho, for the annual Allen & Co. conference, known as 'summer camp for billionaires,' with AI expected to dominate discussions. Leaders from Amazon, Apple, Meta, …
OpenAI's chief futurist Joshua Achiam announced he is leaving the company after nearly nine years, citing a desire to work on the mission from outside the lab. Achiam, who led the mission alignment team and worked at the…
Anthropic's Claude Fable 5 model wrote a booting Windows NT-compatible kernel in Rust in 38 minutes, producing a 5,100-line trusted computing base that passed all self-tests in QEMU. The feat demonstrates AI's ability to…
A developer recreated an AI agent loop in 100 lines of Common Lisp, arguing that the language's recursion and eval capabilities make it ideal for building agents. The project, built with SBCL and two libraries, demonstra…