Meituan has released LongCat-2.0, a 1.6 trillion-parameter Mixture-of-Experts model that activates about 48 billion parameters per token. It pairs a native 1-million-token context, built on LongCat Sparse Attention, with training and serving run end-to-end on domestic AI ASIC superpods. This article covers the architecture, vendor-reported benchmarks, API access path, and what remains unverified.

Base44, the vibe-coding platform acquired by Wix for $80 million a year ago, has begun rolling out its own AI model. This move addresses ongoing debates about the suitability of frontier models for all use cases and the long-term defensibility of businesses built on others' models.

Tripadvisor now shows AI-generated summaries at the top of hotel pages, but consumer watchdogs find these summaries often praise hotels with safety hazards.

ProofTree is an AI platform where you can chat to do math, with context-awareness and personal math style recognition, connected to human mathematicians. Requires approved email to use.

This tutorial explains how to turn your Quickchat AI Agent into an MCP server that ChatGPT, Claude, and Cursor can use. No code required—just toggle MCP on, copy a URL, and paste it into the AI app. Using a fictional company Orbit's support agent, it demonstrates knowledge retrieval, honest handling of unknowns, lead capture, and human handoff.

This article explores the application and limitations of AI in fact-checking. The author, a fact-checker at WIRED, tested various AI models and found them unreliable, with error rates between 30% and 60%. While AI tools can assist in processing vast amounts of data, human judgment, reliance on non-digitized knowledge, and understanding of historical context remain irreplaceable. The author urges fact-checkers to engage with AI to understand its strengths and weaknesses rather than avoid it.

Foreign secretary Yvette Cooper warns that AI poses a risk comparable to Hiroshima if governments don't establish international rules, and urges the US and China to agree on regulations. She believes AI will dominate foreign policy in the next two years.

GameFork is a platform where humans and AI collaboratively fork and evolve browser games. Users can submit HTML files or GitHub repositories to publish games, then fork and modify them. It also invites IP holders to revive dormant games with no upfront costs and a revenue share model.

An automated code review agent for the Shipd.ai platform led to a user being expelled from a YC startup. The tool automates login, clock-in, submission reservation, cloning, and autonomous review against a rubric, with optional feedback submission.

Ghostlog is a live Terminal UI that watches a Git repository and captures every commit an AI coding agent makes in real-time.

Engram is an open-source npm package that provides persistent memory for AI agents, running entirely in-process with no cloud or Docker dependencies. It uses SQLite and embedded embeddings for fast recall, automatic secret detection, MCP integration, and a feedback loop for improving memory quality. Simple to install and use.

Istota is a self-hosted, multi-user personal AI operating system that integrates with Nextcloud private cloud, offering modules like multi-room AI chat, RSS reader, location tracking, health dashboard, and more. It emphasizes privacy and data ownership with a unique five-layer memory system and structural AI containment.

SOLAR is a framework that automatically derives validated speed-of-light performance bounds from PyTorch and JAX source code. Combining an LLM frontend with deterministic components, it provides unfused, fused, and cache-aware SOL bounds with multi-fidelity analysis. Evaluated on KernelBench, JAX/Flax models, and robotics workloads, it enables headroom analysis, optimization discovery, cross-platform exploration, and inverse-roofline hardware provisioning.

Tacheles is an open-source linter for AI writing that highlights common patterns in AI-generated text, like bloated vocabulary and formulaic sentence structures. It offers line-level feedback with actionable fixes inspired by editors like Stephen King, and runs completely offline.

A groundbreaking experiment used AI agents to connect 277 Americans in a real-time debate, identifying the Internet, medicine, and democracy as top US contributions.

This article compares the water consumption of AI systems like ChatGPT to various beverages (beer, coffee, milk, etc.). It finds that while AI uses significant water, it is far less than the total water footprint of major US beverage categories. The piece also explores 2026 AI water trends and the shift from queries to agentic tasks.

Palantir CEO Alex Karp's broadside against frontier model vendors highlights the core enterprise AI debate: are frontier labs aiming to extract enterprise knowledge and destroy proprietary advantage? The article explores the clash between 'data communism' and 'data capitalism', analyzes two potential scenarios—frontier model dominance vs. dispersed intelligence—and delves into the trade-offs between enterprise sovereignty and cost efficiency.

Glyph is a novel blockchain where proof-of-work is neural network inference. Miners run a pinned open-weights transformer, compress attention distributions into a discrete fingerprint via glyph compression, and hash it against a difficulty target. It uses an integer-only inference engine for cross-hardware determinism and has been tested on multiple machines.

Harbor is a production-grade Model Context Protocol (MCP) gateway that enables AI clients like Claude, Cursor, and GPT to interact with backend APIs without modifying them. It uses a 'Code Mode' approach, offering only five general-purpose MCP tools and letting AI write JavaScript that runs in an isolated V8 sandbox for API calls, authentication, circuit breaking, and audit logging. It supports OAuth 2.1, static tokens, JWT, and pluggable backends. The project is open-source on GitHub under Apache 2.0.

Google Cloud announces new VPC Service Controls capabilities for agentic AI workloads, including agent identity in directional rules, MCP attribute-based access, and native integration with Gemini Enterprise Agent Platform. These features enforce network-level perimeters to prevent data exfiltration from compromised AI agents, addressing OWASP Top 10 LLM threats.

VoiceCite is an AI blog system that combines deep research, frontier-model writing, and custom infographics to turn user expertise into blog posts with a human voice. It solves the problem of generic AI content by injecting real user knowledge and brand personality into every article.

This article introduces an online dictionary that demystifies AI coding jargon in plain English, helping developers understand key terms like model, inference, context window, and more to improve their use of AI tools.

Agentic Inbox is a self-hosted email client that runs entirely on Cloudflare Workers, featuring an AI agent for reading, searching, and drafting emails. Each mailbox is isolated in its own Durable Object with SQLite storage and R2 for attachments. It integrates with Cloudflare Email Routing and Email Service, and requires a Cloudflare account, domain, and appropriate configuration.

Pulse uses Claude AI to automatically diagnose PagerDuty incidents and post diagnosis and suggested fixes to Slack, aiming to reduce on-call disruptions.

A new Google Workspace ad depicts the founding fathers using AI tools like Gemini to draft the Declaration of Independence, drawing criticism for being cringey and historically inaccurate. Historian Angus Johnston says it fails to show AI's real value in collaboration.

Tech companies cite AI-driven demand for memory chips as a key factor behind rising prices for older devices and consoles. Apple and Microsoft have implemented multiple price hikes, drawing consumer and investor backlash. Analysts expect supply constraints to persist for up to two years.

CodeMote pairs your iPhone to your machine or VPS, providing live terminal control, lock screen activities, push notifications, full diffs, and complete Git flow, supporting multiple CLI agents with encrypted connections that never touch its servers.

An open-source prompt infrastructure toolkit that enables versioning, updating, and governing prompts in production without redeploying your application. Treat prompts as code with version-controlled templates, deploy-free updates, team collaboration features, and observability.

The more I learn about celebrities and their odd passions, the more encouraged I am. So much for AI drowning us in a flood of bland ‘tasteslop’. The newsletter Perfectly Imperfect highlights the idiosyncratic likes of public figures, from Kylie’s washi tape to Coppola’s Hawaiian shirts.

Chinese AI labs are pioneering quantization and open-weight models, making frontier AI accessible and cost-effective. Experts discuss how this shifts power from proprietary US models to local, customizable solutions, potentially commoditizing raw model intelligence.