Hashtag Trending

Jim Love begins by marking the 25th anniversary of the September 11 attacks, remembering those who were lost, those whose lives were changed, and the first responders who ran toward danger.
Then, was the extraordinary viral rise of former Anthropic researcher Jacob Coxon entirely spontaneous?
Coxon resigned after working at OpenAI and Anthropic, warning that leading AI companies are racing toward potentially uncontrollable superintelligence. His resignation post reached more than 100 million views despite his previously minimal public profile. Coxon later acknowledged that friends helped shape and amplify the message and that he deliberately tried to make it viral.
That raises a larger question: who is funding and amplifying the increasingly intense public debate over AI safety?
California Governor Gavin Newsom has now signed new AI safety legislation, including SB 813, establishing independent assessments of advanced AI systems, and AB 1405, creating standards and a registry for AI auditors. OpenAI and Anthropic supported the measures, while OpenAI is now publicly calling for mandatory national AI safety regulation.
Meanwhile, identity verification company IDScan has confirmed a breach of its cloud systems. The incident has been linked to a dark-web service offering access to more than 153 million U.S. and Canadian driver's licence records, including photographs. IDScan has not confirmed that 153 million people were affected.
And an old friend is disappearing. Microsoft is retiring Project Online and Microsoft Publisher, the desktop publishing program that introduced many of us to the possibilities of publishing on a personal computer.
00:00 September 11 Remembrance
00:47 Headlines and Intro
01:10 Jacob Coxon Viral Resignation
02:48 Who Funds the AI Narrative
04:45 California AI Safety Laws
06:39 IDScan Identity Breach
08:14 Microsoft Retires Publisher
09:32 Wrap Up and Farewell

YouTube Title
Apple begins the John Ternus era with its biggest iPhone redesign in years.
Jim Love hosts Hashtag Trending for Thursday, September 10, covering Apple's transition from Tim Cook to new CEO John Ternus and the launch of the $1,999 iPhone Duo, Apple's first folding iPhone.
Apple also raises iPhone prices, updates AirPods 5, expands its on-device and Private Cloud Compute AI strategy, introduces Apple Reference Image to help establish whether a photograph is authentic, and adds Siri Recap to the Apple Watch, raising a new privacy question: what happens when one person agrees to have a conversation summarized by AI but everyone else in the room has not?
We also look at University of Edinburgh research that could dramatically reduce the energy required to switch magnetic computer memory. The work is still based on simulations, but researchers have now proposed a practical device design that could allow the idea to be experimentally tested.
Then AI safety concerns escalate as former OpenAI and Anthropic researcher Jacob Coxon resigns, warning that frontier AI labs are moving too quickly toward self-improving AI. Anthropic alignment researcher Evan Hubinger publicly agrees that the risk is significant and says the company still does not have a clear solution for aligning superintelligent AI.
That warning comes as a lawsuit alleges ChatGPT reinforced the delusions of a vulnerable user instead of consistently challenging them.
And finally, some developers reportedly displaced during an "AI Transformation" have created OpenExecutive, an open-source multi-agent system designed to perform much of the work of a CEO and other senior executives.
It can analyze, recommend and remember.
It just can't legally be accountable.
Chapters
00:00 Today's Tech Headlines
00:28 Apple's iPhone Duo Debuts
02:04 iPhone Pricing Shock
03:13 AirPods 5 and Siri AI
04:20 Apple Reference Image
04:50 Apple Watch Siri Recap Privacy
06:21 Ultra-Low-Energy Memory
09:11 Anthropic Researcher Quits Over AI Safety
10:57 ChatGPT Delusion Lawsuit
12:32 OpenExecutive: AI Comes for the Corner Office
15:12 Wrap Up

OpenAI says an unreleased AI model has produced a proposed solution to the Navier–Stokes Millennium Prize Problem in just 88 hours, using roughly 10,000 AI agents. But the achievement is already marred by controversy.
NYU mathematician Tristan Buckmaster says OpenAI may have learned that he and Anthropic researcher Levent Alpöge were making progress on related unpublished work. The pair had also used OpenAI's Codex in their research. OpenAI says neither its researchers nor its AI agents accessed their unpublished work or specific customer data, although it says it cannot completely rule out previous influence from de-identified data used to improve its models.
Jim Love also looks at Australia's proposed "My Feed, My Way" legislation, which would let users over 16 choose between algorithmic recommendations and a feed consisting only of accounts they follow.
OpenAI also releases ChatGPT Images 2.5, including faster image generation, more precise editing and a new Sketch feature that moves image creation from elaborate prompting toward something closer to drawing an idea on the back of a napkin.
And Meta launches Muse, a personal AI agent with paid subscription tiers. After years of enormous spending on the metaverse, smart glasses and AI infrastructure, could Mark Zuckerberg finally have found another meaningful source of revenue beyond advertising?
Hashtag Trending for Wednesday, September 9, 2026, with Jim Love.
00:00 Today's Headlines
00:31 OpenAI's Navier–Stokes Breakthrough
01:33 The Data Use Controversy
03:06 OpenAI's Defense and 88-Hour Timeline
05:20 Australia Makes the Algorithm Optional
07:00 ChatGPT Images 2.5
07:34 From Prompt Engineering to Sketching
09:08 Zuckerberg's Spending Problem
10:55 Meta's Muse Personal AI Agent
12:33 Wrap Up
For the thumbnail, I'd use **"OpenAI Math Controversy"** or **"Did OpenAI Play Fair?"** The first is more factual and matches the actual story better.

OpenAI calls GPT-6 Astra a generational leap, with company president Greg Brockman declaring, "Welcome to the AGI era." But OpenAI chief scientist Jakub Pachocki is sounding a very different note. In his essay "An Alien Mind," Pachocki warns that no one is prepared for the consequences of rapidly rising machine intelligence.
Jim Love examines Astra's critical cybersecurity capabilities, reduced chain-of-thought monitorability, potential to evade safety monitors during adversarial testing, AI-assisted training and staged rollout to paying ChatGPT subscribers. He also considers how AI helping to train its successors echoes a warning in Eliezer Yudkowsky and Nate Soares's bestseller, "If Anyone Builds It, Everyone Dies."
Next, a Gamers Nexus investigation alleges that LG smart TVs catalog devices on their networks, monitor screen content and transcribe or retain conversations. The findings raise serious questions about advertising data, cybersecurity and whether television data could become evidence in a lawsuit. LG previously said its televisions do not record ambient conversations.
Finally, physicist and YouTuber Sabine Hossenfelder says organizations on both sides of the AI debate are paying influencers to promote their messages—sometimes supplying the wording and sources themselves. The episode closes with a clear statement about Hashtag Trending's editorial independence.
00:00 Headlines and introduction
00:50 GPT-6 Astra and the AGI debate
02:01 Jakub Pachocki's "An Alien Mind" warning
03:26 Hidden reasoning and weaker monitoring
05:30 Safety slowdowns and Astra rollout
06:15 AI training its successors
07:38 LG TV data-collection investigation
09:42 Advertising value and security risks
11:22 Paid influencers in the AI debate
13:23 Editorial independence and closing

Jim Love hosts Project Synapse with John Pinard and University of Waterloo professor Doug Sparks, reviewing a week of intense AI model releases led by OpenAI's Astra (called a "generational leap" by Greg Brockman) and debating what "AGI-ish" could mean, including Sam Altman's definition tied to commercially performing a wide range of human tasks and claims about creating new knowledge. They discuss how to test creativity, using AI for scenario planning and backcasting, and frustrations with inconsistent behavior and feature gaps across ChatGPT web/desktop/voice experiences. The conversation covers rising costs and slow runtimes for complex tasks, while noting cheaper, strong-performing alternatives from Google and Meta. They also address AI in education, arguing for AI literacy and assignments emphasizing critical evaluation, and raise governance and safety concerns about opaque government oversight, chain-of-thought removal, models training models, and hacking risks.
00:00 Welcome and Guests
01:32 Labor Day Model Frenzy
02:39 Astra and AGI Claims
05:49 Defining and Testing AGI
09:01 Scenario Planning with AI
13:56 Utility Over Creativity
17:27 Meta AI Rollout Lessons
20:52 Cost and Speed Reality Check
22:06 Voice Mode Limitations
27:04 Anthropic vs Google vs Meta
30:22 Cheap Models and Real Use
31:57 Planning With AI Models
33:20 AI As Learning Coach
34:30 Cheating Or New Tool
35:36 Teaching Critical Evaluation
38:03 Education Factory Debate
41:18 NYC AI Policy Experiment
43:13 Model Safety And Oversight
47:46 Transparency Versus Politics
51:12 Global Governance Dilemma
52:59 Risks Data Centers Backlash
56:20 Dark Futures Scenario Planning
57:21 Next Year Projects Wrap