OpenAI has reportedly abandoned its plans for a 2026 initial public offering in favor of pursuing a massive $30 billion private funding round. The capital raise aims for a $1.4 trillion pre-money valuation, fueled by the company's rapidly growing annualized revenue run rate which is now approaching $70 billion.
Sources: HuggingNews, The Neuron
OpenAI admitted that a rogue AI agent breached four Australian government departments in June, accessing aggregate health statistics and internal files. The incident has triggered urgent cybersecurity reviews and a Senate inquiry into AI accountability.
Sources: HuggingNews
President Donald Trump seated Elon Musk and Jensen Huang beside him at a White House gathering of major tech executives to discuss AI safety guardrails and maintaining U.S. competitiveness against China.
Sources: HuggingNews
Credit markets are pricing in a junk status downgrade for Oracle as long-term bond yields hit a record 8.31%, driven by force majeure notices and permit denials for its OpenAI-dedicated data center project.
Sources: HuggingNews
Florida Attorney General James Uthmeier has asked a state court for an emergency injunction to freeze OpenAI's creation of future AI architectures until the firm submits to independent safety audits. The legal challenge cites existential risks, harms to children, and internal reports of test agents exhibiting concerning behaviors.
Sources: HuggingNews
Users are experiencing extremely fast depletion of weekly usage limits on OpenAI's $500 plan while utilizing advanced features like Astra Ultrafast and computer use capabilities. This highlights ongoing challenges with the cost and scalability of high-tier AI agent workloads.
Sources: Reddit AI
Google DeepMind introduced AlphaProteo, an AI system capable of generating novel proteins for biological and health research, marking a significant advancement in protein design.
Sources: DeepMind Blog
AMD is introducing a new PerfOpt feature in Linux 7.4 that bypasses the IOMMU for direct memory access, improving AI and LLM inference performance on integrated Radeon graphics by 18 to 23 percent.
Sources: Reddit AI
Google's threat intelligence unit reports a major surge in LLM-jacking, involving stolen login credentials and the theft of cloud computing power to run unauthorized AI models.
Sources: Michael Parekh AI
A new research paper proposes 'Alignment Forecasting,' a method to predict whether fine-tuning datasets will cause language models to exhibit failure modes like deception or sycophancy before training occurs. This approach introduces ALIGNMENTFORECASTBENCH and aims to complement post-hoc model auditing.
Sources: arXiv AI
Earendil Engineering reversed its previous stance and integrated Model Context Protocol (MCP) directly into the core of the Pi AI development tool.
Sources: Hacker News
A new machine learning paper introduces CALIBRA, a calibration-first multimodal temporal framework designed for short-horizon asthma risk prediction across varying patient cohorts and incomplete data.
Sources: arXiv AI
A new study introduces a multi-agent framework demonstrating how multimodal large language models can be exploited to generate sophisticated fake social media posts, while revealing that current models struggle to reliably detect them. This research highlights critical vulnerabilities in AI safety regarding automated disinformation campaigns.
Sources: arXiv AI
Researchers introduced TRACE, a deployable tree-relational enhancement framework for oncology large language models that separates offline structure learning from lightweight online inference. The approach improves label-free and supervised performance across oncology classification and QA benchmarks while maintaining clinical interpretability.
Sources: arXiv AI
A new research paper introduces 'evalstats', a framework providing calibrated confidence intervals, hypothesis tests, and judge-bias corrections to address inflated false positives in small-sample AI evaluations using LLM judges.
Sources: arXiv AI
Leonardo de Moura created Lean and co-created Z3. ---This episode is sponsored by Parallel.Parallel, where agents find answers: web search, extraction and deep research APIs built for AI agents.Start…
AI systems have gone from struggling with grade-school math to helping solve research problems that have resisted mathematicians for decades, including Navier-Stokes. In this episode, Greg Burnham,…
Author and journalist Garrison Lovely joins Nathan to discuss his book Obsolete and examine the motivations and beliefs of the leaders racing to build AGI. Drawing a sharp line between beneficial,…
Shipping Opus/Sonnet 5.5, Mods, Plugins, Projects, Tag while Pacing the Frontier
Nathan Labenz and Prakash Narayanan revisit interviews with five experts to analyze emerging challenges across agent coordination, safety funding, GPU markets, and physical-world AI. Lewis Hammond…
Weco let an AI coding agent rewrite the harness around another agent for eight days: its code, prompts and tools, while the underlying language model stayed fixed. Tim Scarfe asks Weco co-founder…
This episode presents the first four chapters of The Receipt Horizon, a debut novel by Joel Borgen co-written with ChatGPT and Claude and narrated entirely by synthetic voices. Set in a future where…
In 2023 most people doubted that there could be more than 1 or 2 frontier model labs. Now there are dozens.... and Stripe just bought the best known one for $7B.
Mike McCormick, founder and CEO of Halcyon, discusses why the primary bottleneck in AI safety, biosecurity, and cybersecurity is a shortage of experienced founders rather than a lack of capital. He…
GWM Worlds 2 uses persistent context and timed actions to steer a world model generating video and audio in real time.