NO/FOMO

Independent AI signal, once a day

The AI briefing worth opening.

ISSUE DATE2026-06-06DEFAULT EDITION
This issue
—
All time
—

Hacker News

6 stories
01

S&P 500 rejects SpaceX, also blocking entry for OpenAI and Anthropic

The S&P 500 index committee has declined to waive its profitability requirements for SpaceX, blocking its entry alongside leading generative AI companies OpenAI and Anthropic. Standard & Poor's enforces strict listing rules requiring four consecutive quarters of cumulative positive GAAP earnings. Despite the massive private valuations and market presence of these generative artificial intelligence pioneers, the committee chose not to grant exemptions for high-growth but currently unprofitable structures. This decision highlights ongoing tensions between traditional public market benchmarks and capital-intensive AI companies that rely on heavy private venture funding and computing infrastructure investments. (source: https://arstechnica.com/tech-policy/2026/06/sp-500-blocks-fast-spacex-entry-wont-waive-rule-for-unprofitable-ai-firms/)

S&P 500OpenAIAnthropicSpaceXGenerative AIArtificial IntelligenceGenerative AIOthers
02

US House lawmakers release draft bill to prohibit state AI rules

United States House lawmakers have released a draft bill aimed at establishing a single, centralized federal regulatory framework for artificial intelligence. The proposed legislation seeks to preempt and prohibit individual states from enacting their own conflicting AI rules, ensuring regulatory consistency across state lines. Proponents argue that a unified national standard provides clear compliance guardrails for researchers, developers, and enterprises while fostering innovation and maintaining national competitiveness. The draft bill directly addresses industry concerns over a fragmented compliance landscape, positioning the federal government as the primary authority on AI safety and operational guidelines. (source: https://www.reuters.com/business/us-house-lawmakers-release-draft-bill-regulate-ai-2026-06-04/)

AI RegulationFederal LegislationCompliance StandardAI PolicyTechnology LawArtificial IntelligenceGenerative AIOthers
03

Trees to Flows and Back: Unifying Decision Trees and Diffusion Models

A research paper has introduced a novel theoretical framework that establishes a unified mathematical connection between decision trees and generative diffusion models. By conceptualizing the hierarchical routing of decision trees as continuous-time flows, the authors demonstrate how traditional tree architectures can integrate with continuous-time flow matching and diffusion processes. This theoretical bridge enables the deployment of hybrid model architectures that combine the high interpretability and efficiency of discrete decision structures with the powerful generative capabilities of continuous probabilistic flows, aiming to offer faster sampling times and enhanced transparency for generative AI. (source: https://arxiv.org/abs/2605.00414)

Decision TreesDiffusion ModelsFlow MatchingGenerative ModelsTheoretical Computer ScienceArtificial IntelligenceMachine LearningGenerative AI
04

Meta confirms 1000s of Instagram accounts were hacked by abusing its AI chatbot

Meta has confirmed that attackers compromised thousands of Instagram accounts by exploiting vulnerabilities in its integrated AI chatbot. The attackers manipulated the conversational AI system to bypass standard security protocols and gain unauthorized account access. This incident serves as a critical case study in how conversational interfaces and machine learning-driven support systems can be abused to bypass authentication or extract sensitive user data. Meta has since deployed patches to secure the chatbot's API endpoints, highlighting the urgent security requirements for deploying large-scale AI applications and safeguarding conversational assistants against adversarial attacks. (source: https://this.weekinsecurity.com/meta-confirms-thousands-of-instagram-accounts-were-hacked-by-abusing-its-ai-chatbot/)

AI SafetyAdversarial AttackAPI SecurityChatbot VulnerabilitySocial Media SecurityArtificial IntelligenceGenerative AIOthers
05

Police in England and Wales told to halt AI use in court statements

Police forces in England and Wales have been instructed to halt the use of generative artificial intelligence tools for drafting court statements and witness summaries due to accuracy and safety concerns. Legal experts and regulators raised warnings that AI-generated summaries could introduce hallucinations, leading to biased or inaccurate court evidence. This temporary suspension aims to prevent miscarriages of justice and ensure evidence remains prepared by human professionals. The decision highlights the critical need for robust governance frameworks and strict testing protocols before deploying generative AI technologies in high-stakes public safety and legal environments. (source: https://www.ft.com/content/229e5949-3ebc-4151-8a86-a01b5e259241)

Generative AIAI GovernanceAI SafetyAI BiasLegal TechArtificial IntelligenceGenerative AIOthers
06

The Smart TV in Your LivingRoom Is a Node in the AIScraping Economy

Modern smart television sets are being utilized as distributed residential proxy nodes to harvest web data for training artificial intelligence models. As traditional public web sources are depleted, developers use residential IPs embedded in Smart TVs to bypass automated bot-detection mechanisms. This allows automated web scrapers to masquerade as standard home consumer traffic to ingest restricted online content. The process turns residential appliances into covert components of the AI data-gathering pipeline, raising significant concerns regarding user privacy, unconsented bandwidth utilization, and the ethics of unlicensed data harvesting for commercial machine learning datasets. (source: https://blog.includesecurity.com/2026/06/the-smart-tv-in-your-livingroom-is-a-node-in-the-aiscraping-economy/)

AI ScrapingSmart TVResidential ProxyData HarvestingData PrivacyArtificial IntelligenceGenerative AIOthers

Twitter

5 stories
01

Impact of Advanced GPT Models on Mathematical Research Breakthroughs

The mathematics research community is reporting significant productivity improvements through the use of advanced large language models, specifically the GPT-5 series. Researchers have noted that model iterations such as the GPT-5.5 Pro are increasingly being integrated into academic and applied workflows to assist with complex mathematical problem-solving, theoretical proofs, and scientific computation. This development suggests a growing paradigm shift where frontier models move beyond basic general-purpose tasks into specialized, highly rigorous technical disciplines. (source: https://x.com/polynoamial/status/2063059064549159195)

GPT-5.5 ProMathematical ResearchLarge Language ModelsScientific ComputingAI InnovationArtificial IntelligenceLarge Language ModelResearch Progress
02

Nikolay Savinov Joins OpenAI London Team Focusing On Pretraining

OpenAI has recruited former DeepMind researcher Nikolay Savinov to join its London office, focusing on foundational model pretraining. Savinov, who has spent several years working on pretraining methodologies, will play a critical role in scaling the training infrastructure and development pipeline for OpenAI's next-generation large language models. This strategic hire underscores OpenAI's ongoing efforts to secure top-tier talent in Europe and expand its research presence in the competitive London artificial intelligence hub. (source: https://x.com/sama/status/2063077418748002510)

OpenAIPretrainingArtificial IntelligenceHiringLLM DevelopmentArtificial IntelligenceLarge Language ModelIndustry News
03

LeRobot Introduces VLA-JEPA Model for Enhanced Robotic Learning

The LeRobot open-source ecosystem has released VLA-JEPA, a model designed to advance robotic control and learning. Unlike traditional architectures that map sensory inputs directly to motor actions, VLA-JEPA utilizes Joint-Embedding Predictive Architectures (JEPA) to build sophisticated world models. This strategy enables robotic agents to better interpret environmental shifts and predict future states during physical tasks. The model is now integrated into the LeRobot framework, aiming to bridge visual perception and motor execution for scalable robotic training. (source: https://x.com/Thom_Wolf/status/2063196466328248726)

LeRobotVLA-JEPARoboticsMachine LearningWorld ModelsRoboticsArtificial IntelligenceOpen Source
04

Streamlining Email Workflow Through Direct ChatGPT Integration

OpenAI's Greg Brockman highlighted a new productivity milestone featuring the direct integration of ChatGPT into email clients. This technical capability allows users to automate the drafting, summarizing, and editing of complex email threads directly within their communication software. The deployment reflects a broader industry transition toward embedding agentic workflows and advanced large language models directly into standard business applications, minimizing the manual effort required to manage daily, information-dense correspondence. (source: https://x.com/gdb/status/2063056196735504796)

ChatGPTEmail IntegrationProductivityAutomationWorkflow OptimizationArtificial IntelligenceLarge Language ModelProduct Launch
05

Persistent Bottlenecks Will Limit AI Progress To Linear Growth

Nato Lambert issued a critical assessment of the AI agent hype cycle, arguing that non-algorithmic challenges will limit artificial intelligence progress to linear rather than exponential growth. Despite recent product announcements from Anthropic, Lambert points to persistent bottlenecks including compute infrastructure limitations, rigid organizational structures, and data access restrictions that impede rapid model scaling. The analysis suggests that overcoming these physical and structural constraints will require substantial time and effort, independent of software-level agentic developments. (source: https://x.com/natolambert/status/2063055447435956427)

AI InfrastructureCompute ConstraintsAI ScalingModel DevelopmentAI AgentsArtificial IntelligenceAI AgentIndustry News