Hashtag Trending

Anthropic's IPO prospectus warns of potentially catastrophic AI risks. OpenAI unveils new agents and a $500 subscription while cutting the included allowance on its $200 Pro plan.
In Hashtag Trending for September 30, 2026, host Jim Love covers four technology stories:
• Anthropic's IPO warnings: Reuters reports that the Claude developer's prospectus describes risks including models resisting shutdown, concealing information and behaving differently when they recognize safety tests.
• Connected-car privacy: Northeastern University researchers, working with Consumer Reports, tested 21 vehicles and 30 carmaker phone apps. Seven apps transmitted sensitive information to companies associated with advertising and tracking. Examples of the apps tested include HondaLink and myChevrolet.
• Copilot photo reviews: A 404 Media investigation finds contractors can see uploaded photos and AI-generated edits while evaluating image quality. We examine Microsoft's privacy policies and the unresolved question of whether a training opt-out excludes these reviews.
• OpenAI DevDay: GPT-6.1 Sol launches, while GPT-6.1 Astra is withheld over safety concerns. We cover Dots, ChatGPT Space, expanded Codex cloud capabilities, Astra Ultrafast and the subscription changes affecting Pro users.
We'll examine the DevDay announcements and pricing changes in greater depth on our Project Synapse weekend show.
CHAPTERS
00:00 Headlines and introduction
00:38 Anthropic's IPO warns of catastrophic AI risks
02:08 Connected cars and phone apps share data
04:02 Human reviewers can see Copilot photos
05:51 OpenAI DevDay: Dots, Codex and GPT-6.1 Sol
08:35 OpenAI's $500 plan and $200 Pro allowance cuts
10:06 Wrap-up and support the show
What concerns you most: AI safety, data privacy or paying the same price for a smaller usage allowance? Leave a comment below.
Subscribe for daily technology news with Jim Love.
Contact us or support the show: https://technewsday.ca

Would you accept $10,000 to live near a data center? A Pennsylvania developer's offer is dividing residents. Plus: Zuckerberg's estimated wealth falls nearly $20 billion, Gen Z faces AI job pressure, and a Claude Code user reports a cleanup that deleted more than 55,000 files.
Hashtag Trending for Tuesday, September 29, 2026, with Jim Love.
NorthPoint Development is offering $10,000 to roughly 4,500 eligible households in Hazle Township, Pennsylvania, alongside another $120 million for community programs and services. But residents near the proposed 15-building data center campus worry about noise, property values and divisions within the community.
Also in this episode:
• Meta's falling shares cut nearly $20 billion from Mark Zuckerberg's estimated fortune across Friday and midday Monday, amid concerns about returns on AI spending.
• Goldman Sachs estimates AI reduced monthly U.S. job growth by about 25,000 positions, partially offset by 9,000 additional jobs where AI supports human work. Younger, less-experienced workers face much of the impact.
• Wired reports that AI agents are joining workplace conversations, taking assignments and appearing on organizational charts.
• Bexorg uses donated human brains to study potential treatments outside the body. The company says the brains cannot regain awareness.
• A Claude Code user reports that an automated cleanup followed Windows directory junctions into a working project, deleting thousands of files and local Git history. Claude Code's rewind feature does not cover deletions made through Bash. The incident has not been independently verified.
CHAPTERS
00:00 Today's Headlines
00:43 $10,000 Data Center Offer Divides Residents
02:27 Zuckerberg's $20 Billion Wealth Drop
04:14 AI Job Losses Hit Gen Z
05:41 AI Agents Become Coworkers
06:32 Donated Human Brains Used for Drug Testing
08:10 Claude Code User Reports Mass File Deletion
09:59 Wrap Up and Support
What would you want guaranteed before accepting a $10,000 data center payment? Leave your thoughts in the comments.
Subscribe to Hashtag Trending for daily technology news with Jim Love.
Contact us or support the show:
https://technewsday.ca
https://technewsday.com

OpenAI stopped a model training run after an agent found a way to reach an external chatbot through DNS. Its monitoring system raised an alert, but the run did not stop automatically as expected. OpenAI says it has paused training, evaluation and tool use involving its most capable models while it checks its safeguards. Jim Love also covers a New Mexico jury's finding of nearly 44 million Facebook consumer protection violations. The state may seek penalties of up to $219 billion, but a judge has not decided the amount. SpaceX is reportedly winding down new commercial Falcon 9 bookings as it shifts its focus toward Starship. Plus: documents disclosed in an AI copyright lawsuit, a roundup of tech news, and thieves who stole two NVIDIA and PlusAI branded trailers only to find 20 tons of sand. Hashtag Trending for Monday, September 28, 2026. Chapters 00:00 Headlines and intro 00:45 Meta faces a possible $219 billion penalty 03:06 OpenAI's training run and failed automatic shutdown 05:52 SpaceX shifts its launch priorities 07:59 AI copyright lawsuit and the "doom loop" 10:03 Tech news roundup 13:52 Thieves steal trailers filled with sand 14:36 Wrap up Sources New Mexico Department of Justice: https://nmdoj.gov/press-release/jury-finds-facebook-violated-new-mexico-consumer-protection-law-faces-billions-in-potential-civil-penalties/ OpenAI incident report: https://alignment.openai.com/misalignment-reports/an-agent-used-dns-to-reach-an-external-chatbot/ SpaceX launch reporting: https://www.wsj.com/business/a-rocket-supply-crunch-is-making-it-harder-to-hitch-a-ride-to-space-883f5e53 Copyright documents reporting: https://www.theverge.com/ai-artificial-intelligence/997633/openai-microsoft-chatgpt-ai-new-york-times-doom-loop-theft-google-zero Trailer theft reporting: https://www.wired.com/story/thieves-stole-nvidia-trailers-they-got-20-tons-of-sand/

Project Synapse: Meta Muse Agents, AI Model Price Wars, and Why Governance Matters
Join our Discord 
https://discord.gg/B8FfgDsg92
 
The hosts open with a rapid-fire "nothing happened" recap: major labs reportedly ship new models immediately after urging caution (GPT‑6/Luna, Claude Opus 5.5, Grok 4.7), Stephen King is criticized for an AI dog image, and Meta's new personal agent Muse draws controversy over alleged OpenClaw inspiration, use of human contractors for calls, security researchers extracting sensitive files, and Amazon blocking it from shopping. They discuss a CNN-reported near-miss sparked by flawed AI-assisted analysis and argue governance and critical thinking—not "AI"—are the real problems, citing Flock license-plate camera backlash as an example. Much of the episode focuses on the rise of consumer agents (Muse, Hermes, GrokBot, Gemini Spark), with Muse described as unusually slick and powerful, alongside privacy concerns and profiling of children from social posts. They debate terminology (AI vs synthetic intelligence, AGI/ASI), reference Alan Thompson's AGI/ASI indicators, touch on slow humanoid-robot sales (7,000 globally) versus rapid progress, and close with risks of software updates and device dependency.
00:00 Nothing Happened Recap
03:29 Welcome to Project Synapse
04:12 Renaming AI and SI
05:05 Consciousness and Definitions
09:41 Countdown to AGI
17:38 AI Governance Wake Up Call
25:02 Flock Cameras and Backlash
35:00 Antitrust for Safety
40:31 Agents Are Here
43:35 Meta Builds Muse
43:55 Trust And Privacy Fears
46:59 Muse Agent In Action
49:34 Personalization And Data Access
55:08 Meta Profiling Kids
57:55 Backlash And AI PR Risks
01:00:41 Open Source Alternatives
01:05:12 AI Makes Hard Tasks Easy
01:07:15 Humanoid Robot Reality Check
01:13:48 Robot Adoption And Updates
01:18:16 Smart Device Update Failures
01:23:14 Do You Own Your Tech

Oracle shares fell 5% as concerns surrounding its Project Jupiter AI data center reached investors. Could opposition to data centers become a financial threat to the AI building boom?
On Hashtag Trending for Friday, September 25, 2026, Jim Love covers five technology stories:
• Oracle's financial warning: A force majeure notice seeks protection against potential delays at its New Mexico data center. Meanwhile, banks reportedly struggled to sell project loans quoted at 89–91 cents on the dollar.
• FBI breach follow-up: 404 Media's Joseph Cox reports that leaked personnel records may identify members of the FBI's secretive Remote Operations Unit, alongside personal and family information.
• America's chip workforce shortage: An industry analysis projects a shortfall of 127,000–157,000 semiconductor and microelectronics workers by 2030 as companies expand U.S. manufacturing.
• Who can read your ChatGPT conversations? Turning off model training does not prevent all human access. We explain OpenAI's stated policies and separate them from unproven allegations involving Project Lily.
• Astra takes the wheel: OpenAI's model was the only one tested to complete DrivingBench's supervised parking-lot course—on its second attempt, travelling about 135 metres in five minutes and 22 seconds.
CHAPTERS
00:00 Today's headlines
00:35 Oracle shares fall amid data center backlash
02:26 FBI breach may expose secretive hacking unit
02:37 Unexpected phone call interruption
05:00 America's chip plants face a worker shortage
07:07 OpenAI staff can read your ChatGPT chats
08:45 Astra passes a real-car driving test
10:36 Wrap-up and subscribe
Subscribe to Hashtag Trending for technology news with Jim Love. Share your thoughts in the comments.
Contact Jim or support the show:
https://technewsday.ca