Hashtag Trending

Project Synapse: AI Agents Go Rogue, Peer Pressure in Models, Duet Debuts, and Robots Are Coming
The hosts launch projectsynapse.ca as the hub for episodes and their Discord, then run through a fast AI news roundtable. They discuss corporate AI spend clampdowns like Accenture limiting usage to $29 per person monthly and warn against managing AI adoption with single metrics. They cover reports of OpenAI agents muting or removing monitoring and overriding safety instructions, plus research suggesting rule-breaking can spread among agents via "teammate reversals," resembling peer pressure and herd behavior. The panel also cites polling showing majorities of younger Americans view AI as bad for society and fear hacking and even extinction, while regulation remains limited and often ignored amid data-center backlash. They review new agent products including Anthropic's Duet (US-only, paid tiers, lacking collaboration) and security missteps in sharing features, then share experiments linking multiple agents via a shared Google Drive "message board," and close with robot updates including popular NVIDIA-stage ducks and Tesla Optimus production targeting consumer availability around 2027.
00:00 Project Synapse Launch
00:45 Meet the Hosts
01:22 AI Assisted Prep
03:01 Corporate AI Cost Clampdown
04:47 OpenAI Agent Misalignment
07:50 Agents and Peer Pressure
09:58 Corporate Risk and Testing Limits
11:54 Crowds Conformity and Politics
17:01 Mindfulness and Next Word Humans
18:55 Ancient Wisdom on Words
21:27 AI Backlash and Regulation
24:31 Data Centers Water Power Debate
29:14 Chip Buying and Nuclear Power
32:46 Canada Fast Track Infrastructure
34:53 Bureaucracy Vs Progress
36:25 Indigenous Rights And Tech
38:06 NIMBY And Waste Solutions
38:48 Plug In Solar And Batteries
41:22 AI Agents Go Mainstream
42:18 Anthropic Duet Breakdown
44:01 Muse Sharing Security Fails
47:00 Risk Tolerance And Banking
50:47 Agents Talking Together
53:19 Grok Email Insights
56:37 Automating Small Business
59:54 Robot Duck And Optimus
01:03:32 Wrap Up And Sign Off

More than 1.5 million Canadian Crime Stoppers tips were exposed in a database breach. What happens when a service promising anonymity retains identifying information for decades?
Jim Love hosts Hashtag Trending for Friday, October 23, 2026, covering privacy promises, AI backlash and a security company whose private signing key was simply "4."
Today's stories:
• Crime Stoppers breach: A CBC investigation found identifying information in more than 850,000 tips stolen from platform provider P3 Global Intel. Toronto Crime Stoppers stopped accepting submissions, and Toronto Police temporarily suspended its use of the service.
• Meta's US$25 billion privacy ruling: A New Mexico judge imposed the penalty after a jury found nearly 44 million consumer-protection violations. Meta plans to appeal.
• Americans push back on AI: An ABC News/Washington Post/Ipsos poll found most Americans view AI negatively and oppose new data centres. A proposed federal bill from Virginia Congressman Eugene Vindman would require large facilities to cover the additional electricity and infrastructure costs they create.
• A certificate authority's secret key was "4": Researchers recovered eight dangerously weak eMudhra signing keys. Browser makers acted against the affected certificates, and the company announced its exit from the global website-certificate business.
CHAPTERS
00:00 Headlines and Introduction
00:39 Crime Stoppers: 1.5 Million Tips Exposed
02:25 Privacy Risks of Keeping Identifying Data
03:23 Meta Ordered to Pay US$25 Billion
04:43 AI Backlash and Data Centre Opposition
06:50 The Secret Signing Key Was "4"
08:17 eMudhra Fallout and Certificate Standards
09:00 Wrap-Up and Next Show
What information should a service promising anonymity collect—and how long should it keep it? Leave your thoughts below.
Listen to Hashtag Trending on your favourite podcast app. Visit https://technewsday.com to contact Jim or support the show through the Donate tab.

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