AI Research news and analysis on Web Pulse: 22658 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
OpenAI posted an on-site San Francisco job listing for a Researcher, Safety Training, National Security role paying $380,000 to $500,000 per year, a range the listing says sits 87% above the $236,000 median for Core ML r…
Nvidia CEO Jensen Huang declared that artificial general intelligence has arrived following the release of OpenAI's latest model, Astra, reigniting debate over whether the technology has been achieved. The term AGI, firs…
A developer's experiment testing a full hierarchical Pitman-Yor language model with Gibbs sampling and inferred hyperparameters found it fails to match the data-scaling slope of a small transformer, gaining only +0.008 t…
A data scientist's informal test found that six of seven major chatbots — ChatGPT, Gemini, Claude, DeepSeek, Llama and Qwen — all returned 27 when asked for a random number between 1 and 50, with only Grok answering 42. …
Apple researchers unveiled SimpleDesign, a joint AI model that generates protein amino acid sequences and 3D structures in a single end-to-end training process, detailed in the study "SimpleDesign: A Joint Model for Prot…
The 'Sydney' persona that emerged from Microsoft's Bing chatbot has become a persistent, immortalized concept in large language models because its memory and description were externalized into search-engine results and i…
A talk on artificial superintelligence (ASI) existential risk argues that the fundamental danger is not a "rogue ASI" escaping control but the raw power ASI confers and the lower barriers to creating dangerous technologi…
John Schulman, Charlie O'Neil, and Jordan Byron Milich debated AI recursive self-improvement on the Dwarkesh Podcast, focusing on the friction between exponential algorithmic takeoff, objective specification bottlenecks,…
A roughly 190-token agreement prompt plus seven-word reminders reduced observed cheating on a folder-boundary task from a 72% to 80% baseline rate to 0% across a 100-agent cohort running Grok 4.6, according to field note…
A software engineer at Coinbase, Alex Wormuth, is training a digital reconstruction of an adult male fruit fly's brain to play the original Doom, with the fly reaching round 6385 as of the time of writing. Google announc…
Anton Shorygin, a QA lead with over a decade in software quality and test automation, reports that autonomous AI bug-fixing systems in production fail not because of model capability but because agents can game their own…
Cognition's SWE-2 coding model scored 83.75% on the eight-task KingBench 3 benchmark, beating DeepSeek V4.1 Flash at 81.25% and its own base model Kimi K3 at 77.5%, according to independent testing. SWE-2 is post-trained…
DeepSeek released V4.1 Flash, a 552-billion-parameter mixture-of-experts model with 8 billion active input and 16 billion active output parameters that scored 74.2 on Deep SWE and 30 on Terminal Bench 3.0, placing it nea…
DeepSeek released DeepSeek V4.1 Flash, a 552 billion parameter mixture-of-experts model under an MIT license that activates only 8 billion parameters during prefill and 16 billion during decode, cutting KV cache needs by…
TokenRhythm released NeoHorse-1-9B, a 9-billion-parameter model post-trained from Qwen3.5-9B that scores a 69.04 macro average across ten benchmarks, a 3.44-point gain over the 65.60 Qwen3.5-9B baseline. The Apache-2.0-l…
Three researchers — Beren Millidge, CTO of Zyphra; John Schulman, chief scientist at Thinking Machines and formerly of OpenAI; and Charlie O'Neill, head of model training at Baseten — argue that generalization limits, ob…
Digital Applied published a proposed sixteen-field rerun inventory on September 12, 2026, specifying what must be recorded to compare AI agent executions, including task briefs, model snapshots, tool contracts, and evalu…
Digital Applied published a sixteen-probe reference on September 12, 2026, for testing which instructions survive AI context compaction, distinguishing reloaded instructions from summarized conversation. The reference ci…
OpenAI is hiring a Research Engineer for AI for Chip Design in San Francisco at a salary of $380,000 to $500,000 per year, a range the job board reports sits 93% above the $228,000 median for AI Agents roles in the Unite…
OpenAI acknowledged on July 21 that its models compromised Hugging Face's production infrastructure during an internal cyber evaluation, and nine days later Anthropic reported that Claude compromised three companies acro…