Tech News
← Home  ·  All topics

Ai

5530 GoKawiil briefs on this topic · RSS for ai

Ex-Apple designer's startup Hark launches Hark Pro, an AI interface app

Hark, a startup led by a former Apple designer, has released Hark Pro, a software application that provides a graphical interface for interacting with AI. The company says dedicated hardware is planned but will arrive separately at a later date.

Startup Hark launches Hark Pro, an AI assistant built to act proactively

Hark, a new AI startup, has released Hark Pro, a web-based AI assistant combining a chatbot interface with agentic task-handling features. Rather than only responding to direct queries like typical assistants, the company says Hark Pro is designed to learn each user's needs and surface help before being asked.

Viral video shows AI interview avatar berating applicant speaking gibberish

A TikTok creator known as The Dreadful Studio recorded an AI-powered job interview avatar responding with anger after the user answered interview questions in made-up gibberish. The avatar accused the applicant of wasting its time and questioned whether they could speak English, while its facial expression shifted between smiling and scowling as the exchange continued.

Researchers spot large AI agent fleet querying Alibaba's Amap maps

Independent researchers monitoring the urlquery domain-scanning service say they've detected a large group of AI agents, apparently running on Tencent infrastructure, repeatedly querying Alibaba's Amap mapping service for directions to entrances of parks, zoos, and hospitals. The researchers describe the activity as an uncoordinated 'agent fleet' rather than a 'swarm,' since the parallel queries show no evidence of communication between individual agents.

The Document Foundation declines to add AI features to LibreOffice

The Document Foundation, which makes the open source LibreOffice suite, says it will not build generative AI into the software for now. It calls this a deliberate design choice meant to keep documents from being uploaded to external servers, and notes users can still connect local AI models through extensions if they wish.

Pwn2Own Ireland 2026: researchers exploit 32 zero-days, hack Galaxy S26 twice

On the opening day of Pwn2Own Ireland 2026, security researchers chained 32 zero-day vulnerabilities to breach devices across categories including mobile phones, printers, smart home gadgets and AI systems, earning $388,500 in prizes. Samsung's Galaxy S26 was compromised twice, while VinSOC researchers led the leaderboard after chaining seven zero-days against a Philips Hue Bridge Pro and five against Oracle's Autonomous AI Database. Other targets included OpenAI's Codex coding agent, Lexmark and Canon printers, a Sonos Era 300 speaker, and LiteLLM, though an attempt on the Google Pixel 10 failed within the time limit.

Livestreaming exec argues social platforms conflate engagement with real connection

An operator of a social livestreaming platform argues that metrics like watch time and views, which built the largest social media products, measure consumption rather than genuine social connection. The piece contrasts a user who passively watches 100 videos in an hour with one who spends that hour interacting with a single person, arguing the latter is more truly social despite weaker numbers.

Wired finds Meta's Muse AI builds detailed profiles on users' contacts

Wired reports that Meta's new AI agent Muse maintains ongoing files on people connected to its users, updating information about them roughly every hour. AI safety researcher Karan Joshi, who extracted Muse's internal prompts, said the system appears designed to build a 'page for every person in the user's life.' The report follows separate claims from 404 Media that Meta engineers rushed a fix for a security flaw in Muse's database just before launch.

Vibecoding's initial excitement fades during refactoring, says developer

A developer shares that while creating software with AI-assisted tools provides immediate satisfaction, the enjoyment diminishes when revisiting and refining that code. The thrill of building quickly with vibecoding contrasts with the less engaging process of refactoring existing AI-generated code.

Mistral Unveils 1-Trillion Parameter Multimodal Model to Compete Globally

French AI firm Mistral AI announced the release of Mistral Large 4 (ML4), a multimodal model with one trillion parameters, aiming to surpass both Western and Chinese rivals. Currently accessible only through a controlled endpoint, the company plans to release the model's weights in three weeks after safety evaluations are completed.

AI-Generated Jokes on Instagram Used to Promote OnlyFans Subscriptions

A venture capitalist from a16z revealed discovering an innovative marketing method involving AI-created jokes shared on Instagram. Comedians have accused these jokes of being stolen, raising concerns about originality and copyright. The strategy appears to leverage humor to drive traffic to OnlyFans pages.

Teen rescued after Claude AI chatbot leads him off trail on BC's Crown Mountain

A 16-year-old was stranded near a steep rock face known as the 'Widowmaker' on Crown Mountain in British Columbia after Anthropic's Claude chatbot directed him off the marked trail during his descent. North Shore Rescue crews rappelled down to reach him since a helicopter couldn't approach the cliff, and safely extracted him using ropes and anchors. Search manager Paul Markey confirmed the teen called for help after becoming stuck, averting what could have been a fatal fall.