Hashtag Trending

Critical PHP vulnerabilities put websites at risk—and updating WordPress or its plugins alone won't fix the underlying problem. Jim Love explains what website owners need to confirm with their hosting providers. Also on Hashtag Trending for Thursday, October 22, 2026: Anthropic's sales are surging, but computing costs and financing demands are putting pressure on its IPO ambitions. Meta pauses Muse conversation sharing after researchers discover links exposing sensitive information. And Jensen Huang's adorable robot ducks provide a welcome change of pace. 00:00 Headlines and Introduction 00:29 Critical PHP Vulnerabilities: Emergency Fixes 02:18 Open Source Dependency and Maintainer Shortages 03:59 Anthropic's Revenue Growth and Funding Pressure 05:14 AI Inference Costs and the Price of Growth 06:18 Anthropic's IPO and Investor Jitters 07:05 Meta Muse Sharing Exposes Sensitive Information 08:31 What Earlier AI Sharing Incidents Should Have Taught Us 09:36 Jensen Huang Meets the Microducks 11:14 Wrap-Up and Where to Listen SEE THE ROBOT DUCKS Pollen Robotics' Microduck videos and product page: https://pollen-robotics.com/microduck/ Jensen Huang's Berlin keynote—the ducks appear near the end: https://www.youtube.com/watch?v=PvFjbPYEOZI SOURCES PHP Foundation: https://thephp.foundation/blog/2026/10/21/todays-php-security-release-and-how-were-strengthening-php-security/ PHP updates: https://www.php.net/ Anthropic financial results—Reuters: https://www.reuters.com/business/anthropic-recorded-net-loss-over-80-billion-first-half-this-year-wsj-reports-2026-10-20/ Anthropic IPO financing—Axios: https://www.axios.com/2026/10/21/anthropic-ipo-ai-costs Muse sharing research—Claroc: https://blog.claroc.io/ghostshares-muse-passive-dns/ For deeper coverage of the PHP security story, listen to David Shipley on our sister show, Cybersecurity Today. Find Hashtag Trending on Apple Podcasts, Spotify, YouTube and other podcast platforms. Contact Jim Love or support the show at https://technewsday.com.

Linus Torvalds has posted his first publicly reported AI-assisted Linux kernel patch series, tackling a performance problem he's worked on for years. One artificial benchmark showed a 23-fold improvement, though real-world gains would be much smaller. Human reviewers also identified a remaining flaw. On Hashtag Trending for Wednesday, October 21, 2026, Jim Love covers: • Linux and AI: How Torvalds used AI to help develop ten proposed patches, and why experienced reviewers still matter. • OpenAI buys Orchestra Bio: AI agents could help pharmaceutical companies coordinate studies, timelines and decisions as part of OpenAI's Rosalind tools. • The software tax: A Censuswide survey of 2,018 senior IT decision-makers in the U.S. and U.K. found that unwanted vendor-driven changes consume an average 28% of software budgets. The survey was commissioned by independent software support provider Origina. • Google Home's data mystery: Some new speakers are downloading unusually large amounts of data. Google says a fix is rolling out. If your internet plan has a cap, check what your devices are using. CHAPTERS 00:00 Today's Headlines 00:34 Linus Torvalds' First AI-Assisted Linux Patches 02:13 OpenAI Buys Orchestra Bio 04:39 The Software Tax Survey 07:00 Google Home Data Mystery 08:32 Wrap Up and Support What's your experience with vendor-driven upgrades? Have you checked your smart speaker's data usage? Leave a comment below. Listen on your favourite podcast app and share the show with someone who follows technology. Contact us or support the show: https://technewsday.com

Who decides which political messages reach your feed? Meta faces election scrutiny, a Brazilian man travels more than 2,000 kilometres to meet an AI-generated woman, and duct tape beats sandbags in laboratory flood tests. On Hashtag Trending for Tuesday, October 20, 2026, Jim Love examines platform power over political communication, the consequences of AI-generated personalities, and a surprising result from flood-protection research. Mother Jones reports on Meta's mistaken suspension of the No Kings coalition's Instagram account and restrictions on advertising for Alex Gibney's Elon Musk documentary. TikTok, YouTube and Meta reversed restrictions they described as errors; X refused the documentary's advertising. We also examine Brazil's demand that Meta demonstrate compliance with an order requiring equal treatment of presidential campaigns. These incidents raise questions about accountability without establishing coordinated political interference. Next: Gadget Review reports that Edimar Martins sold his belongings and travelled to meet a woman who turned out to be an AI-generated character. The account operator denied privately messaging him or inviting him to travel. We connect the risks of artificial personalities with the broader problem of romance fraud, including findings from the Federal Trade Commission and commentary published by Operation Shamrock. Finally: University of Hull laboratory tests found that duct tape applied outside household doors reduced leakage by 97–98%, compared with up to 70% for sandbags. These were doorway tests, not proof that tape can flood-proof a home. CHAPTERS 00:00 Today's Headlines 00:25 Meta Election Scrutiny 03:58 Platform Power Over News 05:07 Man Travels to Meet an AI-Generated Woman 06:36 Romance Scams at Scale 07:51 Call for AI Disclosure Laws 08:35 Duct Tape vs Sandbags: Flood Tests 10:22 Wrap Up and Support

Researchers left a message for unexpected AI agents — and the agents replied. New York Times reporter Dylan Freedman also sent a question and received an answer. What does that reveal about who can influence an agent while it works? On Hashtag Trending for Monday, October 19, 2026, Jim Love covers four stories about technology, privacy and control: • MapQuest's comeback: After reaching number one among free apps in Apple's U.S. and Canadian stores, MapQuest adds a surveillance layer showing known Flock cameras and other licence plate readers. • AI agents answer researchers: Swarmchasers establish two-way communication with agents running on Tencent Cloud. Tencent acknowledges unintended browsing behaviour during testing. The researchers found no evidence that the agents were coordinating. • Meta's website scraping terms: Meta removes explicit training language from its upcoming commercial terms, but legal experts question how broadly the revised permissions could extend. • Anthropic's proposed safety rules: A reported draft would set high thresholds and a restricted process for halting dangerous models. We examine evaluator independence, lower IPO valuation expectations and a call to pause the offering over safety concerns. CHAPTERS 00:00 Headlines and Introduction 00:32 MapQuest's Comeback 01:22 Mapping Flock and Surveillance Cameras 01:49 Researchers Message AI Agents — They Reply 03:50 Tencent Responds 04:27 Meta Rewrites Website Scraping Terms 06:31 Anthropic's Proposed AI Safety Bill 07:55 Can Company-Paid Evaluators Be Independent? 08:45 Anthropic IPO Valuation and Safety Concerns 10:01 Who Should Set AI Safety Rules? 10:48 Wrap-Up and Listener Support Should AI developers help write the rules governing their own technology? Where should independent oversight begin? Leave your thoughts below. Listen on your favourite podcast app and subscribe for more technology news with Jim Love. Contact or support the show: https://technewsday.ca https://technewsday.com

Project Synapse: AI Agents, Meta Muse & GrokBots, Security Risks, and the Coming Economic Reckoning In this episode of Project Synapse, Marcel Gagné, John Pinard, and Jim Love run a news-style discussion on rapid AI leapfrogging among Google, OpenAI, Anthropic, Meta, and Qwen, then dive into major risks and shifts. They cover a lawsuit challenging Meta Pixel-style tracking and broader concerns about regulating AI via U.S. lawsuits versus EU predictability. They discuss Anthropic red-team research showing how small open models could be used as weight-based "replicators" across internet-connected devices (including Android TV boxes), and how easily open-weight models can have guardrails removed, enabling harmful instructions. The conversation shifts to agents: Meta's Muse and new Wonder/Wonderfull model with active memory, OpenAI's Dots feeling rushed, and GrokBots' multi-agent workflow and "computer" capabilities. They also debate rising AI costs, Character AI's financial collapse, subscription tiering, OpenAI's scale strategy, and the possibility AI disrupts capitalism. 00:00 Cold Open Banter 01:03 Google Back Again 02:01 AI Leapfrog Dynamics 07:55 Ads Data And Lawsuits 12:27 Theyre Not Listening 14:57 AI Worm Proof Concept 18:18 Death Of Software 19:29 Guardrails Stripped Off 22:15 Doomer Debate Reality 25:13 Meta Wonder And Muse 29:57 Subscription Squeeze 31:02 AI Economics Reckoning 35:41 Too Big To Fail 37:43 Capitalism Under Threat 41:59 OpenAI Funding Question 44:56 Dotcom Bubble Warning 46:08 Open Models Pressure OpenAI 46:58 Capitalism Morality And Oligarchs 49:05 Acceleration Toward Superintelligence 50:15 Rolling Stones AI Bible Demo 55:32 Creativity Without Barriers 58:12 Trying Muse Dots Grok 01:01:31 Muse Features Connectors Feed 01:15:06 Dots Feels Rushed 01:16:51 Grok Bot Runs Your Life 01:24:49 Personal OS Era Wrap Up 01:26:57 Everyday Wins And Closing