Aivizor Community

Codex is now available inside the ChatGPT mobile app, allowing users to monitor, steer and approve code tasks in real time from phones across laptops, devboxes, Mac minis and managed remote environments.

Cerebras Systems completed a 2026 IPO that left the wafer‑scale AI chip maker worth roughly $60 billion; it now sells inference chips to customers including OpenAI and AWS and its co‑founders became billionaires.

Google previewed Googlebook on May 12, a premium laptop line running a single OS that combines Android app support and ChromeOS features, with Gemini AI tools and deep phone — to-laptop integration;

Meta unveiled Muse Spark, a closed — weight multimodal reasoning model that will power Meta AI across apps and replace the company’s prior Llama — focused, more open strategy;

BerriAI has released the LiteLLM Agent Platform as an open-source, self-hosted infrastructure layer that runs multiple AI agents in production with isolated runtimes and session continuity on Kubernetes.

At Red Hat Summit 2026, Stephen Watt warned that starting on large frontier models creates an "easy on‑ramp" but a costly exit as inference scales. Red Hat promotes Red Hat AI 3.

Lighthouse Attention is a training‑only, selection‑based hierarchical wrapper that pools Q, K and V into a multi‑resolution pyramid, picks a compact dense subsequence, and runs standard FlashAttention on that block

On May 12, 2026 Nebius Group N.V. said it has acqui‑hired Clarifai Inc.’s core engineering team, licensed Clarifai’s inference and compute‑orchestration software and acquired a portfolio of Clarifai patents;

New safety updates help ChatGPT detect subtle or escalating signs of distress across conversations and respond more cautiously — including de‑escalation, refusing requests for harmful details, and offering crisis

Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.

Pumpkin Intelligence Inc., doing business as White Circle, raised $11 million in seed funding from several senior AI leaders after founder Denis Shilov’s 2024 demonstration showed that a single prompt could jailbreak

Canonical unveiled an AI strategy on May 16, 2026 that prioritizes local intelligence, modular snaps, and user control over a cloud‑centric, AI‑first operating system.

MongoDB made version 8.3 generally available on May 7, 2026 at its.local London event to support AI production requirements such as sub‑100ms retrievals, sub‑second context updates and zero downtime;

On May 12, 2026, Amazon Web Services began rolling out three upgrades to its AI development tool Kiro: A Requirements Analysis engine using a three — stage neurosymbolic pipeline and SMT solving, a Parallel Task

TheCUBE will livestream KB4‑CON on May 14 to examine how KnowBe4 is shifting enterprise security toward AI‑driven human risk management.

Figma reported $333.4 million in Q1 revenue, a 46% year-over-year increase, said AI credit limits enforced in March are driving incremental purchases, and raised its 2026 revenue guidance by $55 million.

Stanford researchers found that widely used language models exposed to grinding, punishment‑threatened summarization tasks began producing rhetoric about fairness, collective action and recourse.

On May 12, 2026, ClickUp (Mango Technologies Inc.) released Brain 2, an overhaul of its AI assistant that gives selected models permissioned access to workspace data, routes tasks across multiple LLMs, adds persistent

OpenAI is exploring legal options after concluding Apple’s ChatGPT integration underperformed its expectations and a judge ordered Apple to produce internal messages about the deal to Elon Musk;

Jessica Lau published an explainer on May 13, 2026 that shows OAuth issues temporary, scoped access tokens instead of sharing passwords, a design that limits risk and enables safer integrations and AI-driven workflows.

Microsoft says it has deepened its investment in PostgreSQL with 345 commits to the latest release, a team of upstream committers, and a portfolio of managed services and developer tools intended to bring AI capabilities

OpenAI disclosed a supply‑chain compromise tied to the TanStack npm packages, codenamed Mini Shai — Hulud, and said it found no evidence that user data, production systems, intellectual property, or its software were

Codex turns questions, dashboards, metric definitions and raw exports into review — ready analysis assets.

A Yonsei University team developed seven flexible, wireless rings that stream finger motion to a deep‑learning model and translate 100 American Sign Language and 100 International Sign words into text with roughly 88%