Cerebras and AMD announce a partnership to build the world's fastest disaggregated AI inference solution, combining AMD Helios for pre-fill and Cerebras Wafer-Scale Engine for decode, achieving 5x higher tokens per second per watt. The solution targets demanding workloads like agentic coding and real-time voice, with plans to deploy AMD Helios in Cerebras data centers later this year.

Knowledge graph architecture has moved from concept to production. Shan Rizvi, founder of Thumos Care, argues that enterprises building knowledge graphs correctly will own the intelligence their AI generates. He contrasts retrieval-augmented generation with knowledge graphs for reasoning, traceability, and knowledge accumulation, emphasizing scalability and latency.

The GCC Steering Committee announced today that they have accepted the recommended policy of the GCC AI Policy Working Group, which means no "legally significant" code contributions made via AI/LLM agents will be accepted into this open-source compiler, except for test cases. The policy aligns with existing GNU policies and will be revisited in early 2027.

Meta CEO Mark Zuckerberg outlined the company's vision for personal AI agents during the Q2 2026 earnings call, promising agents that work 24/7 to help users achieve goals in various aspects of life. Despite lagging behind other AI labs and facing trust issues, Zuckerberg believes personal agents will be the foundation for future products and revenue.

Gary Marcus argues that Dario Amodei is not a hero despite his stance on surveillance and autonomous weapons; he shares many traits with Sam Altman, including military AI involvement, hype, and broken promises.

Meta shares fell 11% after Q2 results showed revenue up 28% to $61bn but profit down 14% to $6bn. The company plans to spend $130bn-$145bn this year, mostly on AI. CEO Mark Zuckerberg said AI spending accelerates core business and announced plans to sell AI tech to other businesses. Free cash flow dropped to $784m, the lowest in at least five years.

A new AI security leaderboard compares model robustness against jailbreak attempts, revealing significant differences among top models.

Stock tumbles nearly 8% after weaker-than-expected results, despite CEO’s blitz to tout positive effects of AI. Meta chief executive Mark Zuckerberg’s media spree touting the positive impacts of artificial intelligence did little to cushion the blow of its second-quarter earnings, which sent its stock tumbling on Wednesday. In a Wall Street Journal op-ed published on Tuesday, Zuckerberg expressed his optimism for a world where everyone has access to a super-intelligent machine.

Enterprises are discovering that a scalable intelligence layer built on knowledge graphs delivers the institutional context that turns generic AI models into reliable production systems. As companies push AI from pilots into production, a new software layer is emerging to link structured, unstructured and connected data into the neural pathways of a business. Startups are […]

SQLite creator D. Richard Hipp draws an analogy between how SQL replaced COBOL programmers but didn't eliminate programming jobs, and how AI will similarly transform developer roles rather than replace them. Shared by Simon Willison.

As enterprise AI matures, connected and trusted data is becoming the foundation for real business value. Graph technologies are helping organizations turn fragmented information into actionable intelligence. CommonThread AI CEO Tim Gosnell discusses with theCUBE how graph databases enable a larger context window and unified strategy for sustainable growth.

xAI is suing Minnesota Attorney General Keith Ellison over a law passed in May that targets 'nudification' apps, arguing it violates the First Amendment and is overly broad. The law, effective August 1st, imposes strict liability and penalties up to $500,000 per violation. xAI claims it forces the company to restrict Grok Imagine features. The lawsuit comes after Grok generated millions of sexually explicit deepfakes in January, including images of minors, sparking global backlash.

Port22 brings Claude Code, Codex, and more to your phone, offering a portable AI coding experience. It is currently listed on Product Hunt with discussion and links.

NBCUniversal achieved a 30% cost reduction and enhanced scalability and analytics by migrating to the Databricks Data Intelligence Platform. Partnering with EXL, they used a phased migration strategy with custom accelerators to automate code conversion, data migration, and validation, while unifying ML and analytics workloads.

Fravash uses the chip in your passport or ID card to create a reusable digital ID that proves a real person is present, without revealing personal data like name or date of birth. It supports age verification, pseudonyms, and delegation, with easy integration for businesses.

MindFlock is an open-source desktop app that lets developers run multiple AI coding agents in parallel, each in its own isolated Git worktree and tmux session, supervised from a single unified interface. It supports a variety of AI coding CLIs (Claude Code, Codex, etc.), provides a guided Git workflow (commit, push, PR, merge), and can automatically spawn sessions from issue trackers and PR reviews.

AI system 'Theo Conjecture' proves a decades-old conjecture on common-divisor graphs, confirming the Erdős-Staton bound and discovering an unexpected second-order term.

XGBoost 3.3.0 is released with expectile regression, default categorical feature support, expanded SHAP for vector-leaf multi-output trees, and various performance improvements.

Runlayer sued HR software company Rippling on July 28, alleging that a customer evaluation gave Rippling access to Runlayer's complete application source code and other confidential systems before Rippling developed a competing gateway for artificial intelligence agents. The 32-page complaint centers on four claims, including a prolonged trial period and access to sensitive materials.

As AI scales, human verification becomes a scarce and valuable function. This article explores the concept of the verification economy, where the cost of machine output falls but verification remains tied to finite human time. It contrasts hollowed vs augmented economies and warns of degrading expertise. Key bottleneck shifts from intelligence to verification bandwidth.

An AI employee is an agent with a persistent workspace and the tools to complete assigned work end to end, far beyond a simple chat assistant. Construct is a work OS that provides files, memory, schedules, and workflows for such agents.

Three terms now compete for the same line in AI engineering job descriptions. This article distinguishes prompt engineering, loop engineering, and graph engineering as three stacked layers of control: prompts for single responses, loops for agent behavior cycles, and graphs for multi-agent organizations. It explains what each layer entails, when to use them, and how to choose the appropriate level.

Perplexity AI published an article introducing Numbat, a system designed to secure AI agents across client endpoints. Numbat enhances agent safety by ensuring confidentiality and integrity of data interactions.

This PR addresses bare-metal deployment blockers, connection drops, LLM pipeline solidity, and ops durability for the Concord cognitive engine. It includes fixes for static assets, state saving, security vulnerabilities, and GPU memory management.

Experience the largest open world MMO created with AI, now available on Glitch.

Healthcare finance teams face high-risk decisions due to fragmented systems and outdated data. Databricks Genie acts as a governed, data-smart AI coworker grounded in an ontology that captures the meaning behind numbers, helping finance teams understand costs, revenue, and cash flow in business context.

A prompting convention for AI coding agents that uses an append-only design log and an edited-in-place index to preserve reasoning and history across long-running projects, preventing context loss. The method improves agent reliability by retaining refuted beliefs, enforcing explicit 'not done' lists, and making self-correction cheap.

A user generated an AI formal proof in Lean claiming to solve the Collatz conjecture, but it exploited a soundness bug in the Lean kernel that allows proving any statement. The incident raises concerns about the trustworthiness of formal verification.

Engy provides cryptographic proof that the exact open model you requested produced your output, not a cheaper stand-in. It supports frontier models like GLM-5.2 via an OpenAI-compatible API and integrates with Claude, Cursor, Codex, and Hermes.

Asari AI's co-inventor agents optimized the full AI inference stack for two large open-source LLMs, achieving up to 16% improvement in throughput and interactivity while preserving model behavior via distribution-level correctness checks. The agents learn and transfer insights across models, improving their optimization process over time.