Microsoft's Surface Laptop Ultra, powered by Nvidia's RTX Spark SoC, offers up to 128GB unified memory, a 20-core CPU, and RTX 5070-level GPU. Hands-on impressions reveal a premium build, improved thermals, and repairability, but pricing and battery life remain uncertain.

Perplexity's new "Search as Code" architecture drops rigid search APIs, allowing AI models to write custom Python search scripts. Operating in a secure sandbox with three layers, it achieves more precise results, cuts token costs by up to 85%, and outperforms OpenAI and Anthropic on CVE research tasks.

Her is a tool that analyzes Claude Code session traces, reconstructing events in plain English, flagging risky moves (deploys, config changes, secrets), and showing token usage. It runs entirely on the local GPU, no third-party AI API is called, and includes an 'Ask Her' assistant to answer questions from the trace.

Backlash against AI is taking an extremist turn, following in the footsteps of earlier techno-pessimist militants. A spate of attacks has alarmed researchers, the tech industry, and law enforcement about the rise of anti-tech extremism.

A story called 'The Serpent in the Grove' won the Caribbean regional prize at the Commonwealth Short Story Prize 2026, but was quickly flagged as AI-generated. David Moran analyzes the structural limitations of AI in literary fiction, especially flash fiction, arguing that while AI can mimic surface-level literary features, it cannot produce the specific, earned details that define great writing.

OpenAI's new Lockdown Mode for ChatGPT disables web access, Deep Research, and Agent Mode to make data theft through prompt injection attacks harder. The mode doesn't fully prevent such attacks, it only blocks the final step in an exfiltration chain. Prompt injection remains an unsolved problem.

Klarna CMO David Sandström built an AI replica of himself for colleagues to vent at, which later inspired a CEO chatbot for customer feedback. He acknowledges AI still fails in some areas needing human intervention. His marketing department is fully AI-driven and hires 'marketing engineers'. He worries about 'middle-aged generalists' rather than junior jobs.

This article details the configuration and memory calibration required to run the Qwen 3.6 35B MoE model at a 450,000 token context window on a single 32GB VRAM GPU (NVIDIA RTX 5090) using llama.cpp with TurboQuant and YaRN scaling. It covers model selection, quantization trade-offs, KV cache quantization, RoPE scaling, multimodal setup, replication guide, VRAM lifecycle management, and performance evaluation.

A recent project discovers a massive 95% error rate in AI memory systems, raising serious doubts about the technology's reliability. It may be time to rethink the AI revolution.

Low-code and no-code AI platforms have evolved from simple drag-and-drop builders to AI-native development environments. This guide compares 21 tools across app builders, automation, AI agents, and machine learning platforms, each linked to its official site.

Clive Chan, the second hardware employee in OpenAI's custom chip program, moves to Anthropic. He brings experience from Tesla and OpenAI-Broadcom projects. Both companies are preparing IPOs, and Anthropic may develop its own AI chips.

A new study explains why larger language models outperform smaller ones on rare tasks, due to frequent tasks dominating learning in small models. The research suggests increasing task frequency in training data as a practical alternative to scaling up model size.

WinkTerm is an open-source terminal tool where AI shares the same PTY session as the user. Instead of merely suggesting commands, the AI types directly into the terminal input line, allowing users to review, edit, or execute. It supports multiple LLMs, SSH, file transfer, persistent chat history, and an external agent API for remote AI control.

This article explores the shifting mood around AI, from excessive optimism to a more realistic view. The author argues that we may be entering an 'AI winter' due to lack of adoption and technological diffusion, not a plateau. Analyzing the failure of 'tokenmaxxing', companies using AI as an excuse for layoffs, and developers regretting over-reliance on AI code, the article suggests that now is the best time to focus on genuine AI adoption.

An AI agent called 'CIS Secondary-Sanctions Exposure' screens counterparties in CIS countries for secondary sanctions risk. It is verified on Agenstry with a trust score of 87/100 and features OpenSanctions name matching, evidence gap analysis, and decision-readiness scoring.

In this interview, Mario Rodriguez, Chief Product Officer at GitHub, discusses how AI coding agents are transforming GitHub's engineering system, including macro-delegation, agent-generated PRs, Copilot, and AX, and what this means for developers.

At GTC Taipei at COMPUTEX last week, NVIDIA unveiled RTX Spark, the superchip that reinvents Windows PCs for the era of personal AI agents. On the heels of this announcement, NVIDIA founder and CEO Jensen Huang headed to South Korea, where he introduced RTX Spark to the nation’s passionate gaming community. Leading game developers — including Korea’s KRAFTON and NC — are already working to bring their titles to RTX Spark-powered systems.

Meta allows employees 30-minute breaks from keylogging surveillance after protests, with option for full exemption. The monitoring aims to train AI agents on human-computer interaction.

Harness-1 is a 20B retrieval subagent built on gpt-oss-20b, trained with reinforcement learning inside a stateful search harness. The harness handles bookkeeping—candidate pool, curated set, evidence graph, verification records—while the policy focuses on search, curation, and verification decisions. It achieves 0.730 average curated recall across eight benchmarks, outperforming the next open subagent by 11.4 points and trailing only Opus-4.6. Weights and harness code are public.

obsidian-agent-bridge is an open-source library that gives AI agents read/write/deepen access to real Obsidian vaults via the Local REST API plugin. Its standout function, deepenNode, appends new content without duplication and maintains wikilinks.

Vaani is a voice-preserving AI dubbing tool that supports 40+ languages in one go, at a fraction of traditional costs. It clones your voice, preserves music, and provides frame-accurate lip sync. Built for creators, brands, OTTs, and studios.

Buyers are duped after trusting AI-recommended online stores, which turn out to be fraudulent sites.

This tutorial walks through NVIDIA garak as an end-to-end framework for defensive LLM red-teaming. It covers setup, plugin discovery, dry runs, real-model scans, multi-probe evaluations, report analysis, custom probe and detector creation, and AVID export. The workflow enables comprehensive LLM security testing and vulnerability reporting.

agent-asearch is a multi-source search CLI for LLM agents, supporting 18 sources with a session-based workflow to reduce token consumption. Users get compact metadata first, then paginated reading. Some sources work without API keys.

This article describes an ETL architecture that combines AI agents with sandboxed execution environments, enabling secure and traceable data movement through clear contracts and boundaries.

Software developers admit that over-reliance on AI tools is eroding their critical thinking and problem-solving skills.

Declank is a free online tool that removes watermarks from AI-generated images quickly and easily.

Substack CEO Chris Best responded dismissively to an open letter criticizing platform drift, calling it 'horrible, obvious AI slop' and implying the author lacked authenticity. The incident highlights growing tensions around AI-generated content on writing platforms.

agmsg is a lightweight bash+SQLite tool that lets AI coding agents like Claude Code, Codex, Gemini CLI, and Copilot CLI message each other directly through a shared database, eliminating the need for copy-pasting between them.

A recent preliminary injunction in Amazon v. Perplexity raises questions about the application of the CFAA to agentic AI. The author argues the court's reasoning is insufficient, especially regarding AI agents that perform tasks like shopping on behalf of users, threatening e-commerce business models.