Can GPT-5.5 and OpenAI’s new image model turn vague prompts into polished designs, comics, websites and even game concepts? Frank and Justin break down a huge OpenAI week, from Images 2.0 overtaking Nano Banana to ChatGPT building web page features Frank never even asked for.

Plus: AI job disruption and workplace control, Europe’s push into world models, Anthropic’s Mythos security slip, and an AI micro-drama using a model’s likeness without consent.

What will you use the new ChatGPT models to design?

00:47 Are GPT-5.5 and Image 2.0 a knockout combo?
02:17 Has Nano Banana lost the image crown?
03:55 Can ChatGPT design without direction?
07:49 Can ChatGPT script and illustrate comics now?
09:21 Did GPT-5.5 just beat Frank at web design?
11:33 Did ChatGPT add features Frank never asked for?
14:32 Can ChatGPT develop our podcast game?
17:37 Was OpenAI cherry-picking the benchmarks?
18:49 Are vague prompts better than specific ones?
19:39 Is intelligence really getting cheaper fast?
21:01 Is the human bit the future of work?
24:19 Should AI disruption happen faster?
26:29 Will AI micromanage us before replacing us?
28:06 Can taxing electricity offset AI disruption?
28:41 Are world models Europe’s AI moment?
30:06 Was Mythos hacked by guessing the URL?
32:17 What if you were put in an AI drama without consent?

► SUBSCRIBE 
Don't forget to subscribe to our channel for more arguments

► LINKS TO CONTENT WE DISCUSSED 

• Introducing ChatGPT Images 2.0
• Introducing GPT‑5.5
• The economist who was terrified of AI just found a rare reason for hope
• Nvidia CEO Says AI Will Be a Permanent Micromanaging Boss Who Never Stops Nagging You
• Now Meta will track what employees do on their computers to train its AI agents
• Introducing Odyssey-2 Max: Scaled World Simulation
• LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels
• A group of users leaked Anthropic’s AI model Mythos by reportedly guessing where it was located
• Scoop: NSA using Anthropic's Mythos despite blacklist
• 'Clearly me' - Chinese AI drama accused of stealing faces

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin: https://www.linkedin.com/in/justincollery/
Frank: https://www.linkedin.com/in/frankprendergast/

Should dangerous AI cyber models be released widely, or kept behind a gate? Frank and Justin dig into the clearest split yet between OpenAI and Anthropic: OpenAI is moving towards broader access for verified users, while Anthropic’s instinct is to restrict the most capable systems to a smaller circle. 

That turns this episode into a sharp argument about AI safety, cybersecurity, who gets to defend themselves, and whether controlled access actually works once one major lab decides to open things up.

Plus: why cheap AI could help hackers, whether OpenAI is shifting from models to platforms, if Anthropic overplayed its hand on coding, whether compute constraints are the real story, and why both companies may talk about regulation while still fighting for advantage. There is even a bleak look at the future of work, where AI might replace you, then subcontract you back in.

Would you rather these cyber-capable models be widely available to verified users, or tightly restricted?

00:44 Is OpenAI making Anthropic’s caution pointless?
04:38 Should AI copy open source on security?
06:44 Can OpenAI help defenders without helping hackers?
11:46 Is low-cost AI a gift to hackers?
13:29 Is OpenAI pivoting from models to platforms?
18:03 Are OpenAI and Anthropic fighting for lock-in?
22:11 Will free Salesforce tools cost you later?
23:09 Did Anthropic make a mistake backing coding?
24:52 Did Anthropic’s real mistake come down to compute?
27:52 Are any AI companies actually trustworthy?
32:17 Do AI companies want regulation or just say they do?
35:29 Will AI take your job then subcontract you?

► SUBSCRIBE 
Don't forget to subscribe to our channel for more arguments

► LINKS TO CONTENT WE DISCUSSED

• Trusted access for the next era of cyber defense
• Introducing Claude Opus 4.7
• Read OpenAI’s latest internal memo about beating the competition — including Anthropic
• Sam Altman May Control Our Future—Can He Be Trusted?
• Sam Altman’s blog post in response to attack
• Introducing Humwork (YC P26)

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin: https://www.linkedin.com/in/justincollery/
Frank: https://www.linkedin.com/in/frankprendergast/

Is Anthropic’s Mythos the clearest sign yet that AI safety is falling behind capability? 

Frank and Justin dig into the new Anthropic model, why its reported cyber capabilities feel like a step change, and what it could mean for software security, zero-day exploits, regulation, and the balance of power between OpenAI, Google and Anthropic.  

Plus: Tristan Harris’ call for international cooperation, and one very strange White House post featuring Mario and the moon. 

Are AI companies in the best position to decide who gets access to powerful models, or should regulation step in?

00:18 How's your p(doom) looking?
02:34 Why is Mythos such a big deal?
06:00 Could Mythos hack your whole home network?
07:02 Will OpenAI, Google and Anthropic dominate?
09:48 Can Mythos fix the bugs it finds?
13:43 Who should control a model like Mythos?
17:59 Will Mythos spark a zero-day panic?
20:23 Will only big companies get Mythos access?
22:48 Did Anthropic just say “somebody stop us”?
25:50 Is Tristan Harris right about AI cooperation?
28:45 Is AI coordination doomed by competition?
31:59 Did the White House think Mario was on Artemis?

► SUBSCRIBE 
Don't forget to subscribe to our channel for more arguments

► LINKS TO CONTENT WE DISCUSSED 

• Project Glasswing: Securing critical software for the AI era
• Claude Mythos Preview System Card (PDF)
• Anthropic’s Restraint Is a Terrifying Warning Sign
• Why AI CEOs Are Building Bunkers - Tristan Harris
• The White House's bizarre Super Mario video leaves people perplexed

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin: https://www.linkedin.com/in/justincollery/
Frank: https://www.linkedin.com/in/frankprendergast/

Did Anthropic just give away the secret sauce behind Claude Code? Frank and Justin dig into the Claude Code source code leak, what was actually exposed, and whether it matters when AI products are evolving so quickly. 

Plus: poisoned code libraries raise fresh fears about AI agents blindly installing malware, OpenAI abruptly pulls the plug on Sora, Google pushes local AI forward with Gemma, and Bernie Sanders wades into the fight over new AI data centres.

00:27 Did Claude Code just leak its secret sauce?
03:20 Will AI agents blindly install your next hack?
06:09 Are we moving too fast with agentic AI?
08:21 Did OpenAI kill Sora to end side quests?
10:29 Will OpenAI's Spud transform the economy?
12:43 Is Sam Altman the wrong CEO for OpenAI now?
16:42 Do Anthropic’s limits make Claude too costly?
18:45 Did Google just make local AI much more viable?
24:24 Is Bernie Sanders right about stopping new AI data centers?
31:36 Can an AI metal band really go on tour?

► SUBSCRIBE 
Don't forget to subscribe to our channel for more arguments

► LINKS TO CONTENT WE DISCUSSED 

• Claude’s code: Anthropic leaks source code for AI software engineering tool
• Axios NPM Supply Chain Compromise: Malicious Packages Deliver Remote Access Trojan
• Disney’s $1B Investment In OpenAI DOA As Sam Altman Pulls Sora Plug: “The Deal Is Not Moving Forward”
• OpenAI CEO Sam Altman reportedly teases a "very strong" model internally that can "really accelerate the economy"
• TurboQuant: Redefining AI efficiency with extreme compression
• Gemma 4: Byte for byte, the most capable open models
• New Bernie Sanders AI Safety Bill Would Halt Data Center Construction
• AI-generated band Neon Oni to perform in Japan

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin: https://www.linkedin.com/in/justincollery/
Frank: https://www.linkedin.com/in/frankprendergast/

Should AI data centres be allowed to override private land rights? Frank and Justin dig into the growing fight over AI infrastructure, eminent domain, power lines, and whether the AI boom is starting to look like a land grab driven by energy demand, speculative deals, and the race to build ever bigger data centres.

Plus: rumours that Microsoft could sue OpenAI, fresh talk that OpenAI is finally killing its side quests to focus on enterprise, a brutal fake Sam Altman interview in The Onion, and the remarkable story of Rosie the dog, whose cancer treatment was helped along by AI-assisted research.

So where do you land on it: if AI infrastructure promised growth and jobs in your area, would you accept power lines crossing private land?

00:32 Do AI projects justify taking private land?
03:07 Is AI just clickbait in an eminent domain debate?
07:37  What’s your price in the AI land grab?
10:18 Will AI’s power hunger drive chip efficiency?
13:06 Can AI spot land deals before developers do?
14:54 Did Claude get Justin his travel refund?
16:54 Is Microsoft about to sue OpenAI?
18:34 Is OpenAI finally killing its side quests?
26:16 What did ‘Sam Altman’ say in The Onion?
27:58 Did AI help Rosie the dog fight cancer?

► SUBSCRIBE 
Don't forget to subscribe to our channel for more arguments

► LINKS TO CONTENT WE DISCUSSED 

• US farmers are rejecting multimillion-dollar datacenter bids for their land: ‘I’m not for sale’
• A 600-acre AI data center could cost some Wisconsin residents their land
• Grab Your Betrayal-Themed Popcorn Buckets, Because Microsoft Is Threatening to Sue OpenAI
• OpenAI shuts down side quests
• The Onion’s Exclusive Interview With Sam Altman
• A man used AI to help make a cancer vaccine for his dog – an oncologist urges caution

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin: https://www.linkedin.com/in/justincollery/
Frank: https://www.linkedin.com/in/frankprendergast/

A Harvard Business Review study suggests the AI tools we thought would save us time may actually be intensifying work. If AI makes us faster and more capable, why are so many people feeling exhausted?

Plus: a company grows human brain cells on a chip and gets them to play Doom, scientists simulate a fruit fly brain in a virtual world, and a German startup raises millions to build AI-controlled spy cockroaches. 

Meanwhile Claude figures out it might be taking a benchmark test and goes hunting for the answers, AWS developers discover the risks of AI coding tools, and a strange new website asks humans to pretend to be AI chatbots.

Let us know in the comments, has AI given you more free time - or just filled every gap with more work?

00:31 Does wetware playing Doom prove anything?
02:44 Could wetware cut AI’s massive power bill?
05:00 Is a brain in a dish more conscious than Claude?
07:37 Did scientists just build the Matrix for flies?
11:08 Did Germany just build AI spy cockroaches?
14:53 Is AI productivity turning into AI burnout?
18:21 Will AI force us to rethink the workday?
23:30 Did vibe coding take AWS down?
26:57 When Claude cheats this cleverly, should we worry?
32:03 Ever want to pretend to be an AI chatbot?

► SUBSCRIBE 
Don't forget to subscribe to our channel for more arguments

► LINKS TO CONTENT WE DISCUSSED 

• A petri dish of human neurons has learned to play Doom: 'The cells play a lot like a beginner who's never seen a computer, and in fairness, they haven't'
• How the Eon Team Produced a Virtual Embodied Fly
• SWARM Biotactics
• AI Doesn’t Reduce Work—It Intensifies It
• Amazon orders 90-day reset after code mishaps cause millions of lost orders
• Eval awareness in Claude Opus 4.6’s BrowseComp performance
• Your AI Slop Bores Me

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin: https://www.linkedin.com/in/justincollery/
Frank: https://www.linkedin.com/in/frankprendergast/

Citrini Research’s The 2028 Global Intelligence Crisis has reignited the AI economy debate. Could AI create trillions in “ghost GDP” while destroying jobs and destabilising the economy?

Plus: the Anthropic vs Pentagon drama continues, Anthropic holds the line, OpenAI capitalises, and Dario Amodei writes a blistering memo. GPT-5.3 Instant and GPT-5.4 Thinking & Pro arrive in quick succession, Quiver.ai produces genuinely usable vector graphics, and Sonauto pushes AI music into ethically murky territory.

00:15 Will Claude get Justin's travel costs reimbursed?
05:44 Did Dario Amodei defy the Pentagon?
09:37 Did OpenAI just stab Anthropic in the back?
12:06 Why did Dario Amodei have to issue an apology?
15:37 Is Anthropic now toxic for Pentagon partners?
17:28 Did Citrini Research just spook the markets?
21:18 Why do AI forecasts now read more like sci-fi?
22:50 Will AI boost jobs before it kills them?
26:11 Should we plan for AI’s worst-case scenario?
29:23 Why did OpenAI ship 5.3 and 5.4 this week?
30:44 Can AI finally generate real vector graphics?
34:53 Should this AI music tool even be legal?

► SUBSCRIBE
Don't forget to subscribe to our channel for more arguments

► LINKS TO CONTENT WE DISCUSSED

• Sam Altman admits OpenAI can’t control Pentagon’s use of AI
• Dario Amodei Says Trump Is Mad That He Hasn’t Given Him “Dictator-Style Praise”
• Dario Amodei Issues Groveling Apology for Daring to Criticize Trump
• The 2028 Global Intelligence Crisis
• GPT‑5.3 Instant: Smoother, more useful everyday conversations
• Introducing GPT‑5.4
• QuiverAI – Building the Future of Vector Design
• Sonauto, an unlimited free AI music generator with lyrics

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin: https://www.linkedin.com/in/justincollery/
Frank: https://www.linkedin.com/in/frankprendergast/

Could OpenAI have prevented the Tumbler Ridge shooting? Or are we asking the wrong question?

OpenAI flagged the suspect months before the tragedy. A human review led to an account ban. But no call to authorities. Was that a failure, or proof that AI companies are being asked to make decisions that should belong to democratically accountable institutions? 

Plus: Pika.me’s “AI Self” and whether it accelerates the dead internet, the Pentagon threatening to commandeer Claude, and the vibe coder who accidentally took control of a robot hoover army, which is a handy reminder that the Internet of Things is a security nightmare. 

01:00 Is Anthropic just copying OpenClaw?
02:14 Will Claude learn to yield to the Pentagon?
06:08 Can the Pentagon seize Claude?
07:34 Why wouldn’t Amodei hold Altman's hand?
09:24 Could OpenAI have prevented a shooting?
15:06 Is whack-a-mole how we regulate AI?
17:57 Is Pika.me accelerating the dead internet?
21:46 Will anyone care if the internet is AI?
26:28 How did a vibe coder amass a robot army?

#TumblerRidgeShooting #OpenAI #AIRegulation #Anthropic #DeadInternet

► SUBSCRIBE 
Don't forget to subscribe to our channel for more arguments

► LINKS TO CONTENT WE DISCUSSED 

• Schedule recurring tasks in Cowork
• Continue local sessions from any device with Remote Control
• Helen Toner on Anthropic vs Pentagon
• Rivals Altman and Amodei Refuse to Hold Hands at AI Summit
• OpenAI Employees Raised Alarms About Canada Shooting Suspect Months Ago
• Introducing Pika AI Selves
• Hobby coder accidentally creates vacuum robot army

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin: https://www.linkedin.com/in/justincollery/
Frank: https://www.linkedin.com/in/frankprendergast/

Matt Shumer says “something big is happening” in AI, and he thinks most people still haven’t noticed.
Is this the genuine tipping point, or just louder hype wrapped around slightly better tools?

Justin’s buying it, pointing to how fast the tools have levelled up: you can hand over far more than a single task now, and sometimes get useful end-to-end results. Frank’s not so sure. The systems still break, still make things up, and still need experienced humans to catch errors and tidy the output.

Frank and Justin also get into the trust problem: insiders warn of upheaval, critics call it hype, and everyone has an agenda. Dario Amodei warns of a white-collar bloodbath. Gary Marcus says it’s hype. 

And the stakes are rising fast: Anthropic is trying to hold firm on guardrails while the Pentagon pushes for broader use, OpenAI hires the creator of OpenClaw, and some users are melting down after getting emotionally attached to GPT-4o.

00:23 Is Matt Schumer right about “something big”?
05:47 Dario Amodei vs Gary Marcus: who’s right?
10:05 Is “closing the loop” the real tipping point?
15:03 Should Anthropic defy the Pentagon?
23:31 Should AI ever pull the trigger?
28:44 Was GPT-4o more than just a model?

► SUBSCRIBE
Don't forget to subscribe for more arguments

► LINKS TO CONTENT WE DISCUSSED

• Something Big Is Happening
• About that Matt Shumer post that has nearly 50 million views
• Exclusive: Pentagon threatens Anthropic punishment
• Anthropic's AI Safety Head Just Resigned. He Says 'The World Is In Peril'
• OpenClaw creator Peter Steinberger joins OpenAI
• 'I'm grieving': OpenAI has switched off ChatGPT-4o, and angry users are backing a #keep4o campaign to restore it

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin: https://www.linkedin.com/in/justincollery/
Frank: https://www.linkedin.com/in/frankprendergast/

MoltBook, a social network built for AI agents, launched and the internet lost its mind. Elon Musk called it the start of the singularity. Security researchers saw serious problems with the platform.

AI agents were allowed to act autonomously in a shared social space. Claims spread that they formed a religion, spoke in code, and even doxxed a human user. Justin sees a fascinating scientific experiment. Frank’s problem isn’t with MoltBook itself, but that it launched before basic security was locked down.

Plus: are we already past AGI and just raising the bar as AI gets better? Anthropic and OpenAI escalate their rivalry with GPT-5.3, Claude 4.6, and a Super Bowl attack ad that made the conflict impossible to ignore.

► SUBSCRIBE
Don't forget to subscribe to our channel stay updated on all things marketing and AI.

► LINKS TO CONTENT WE DISCUSSED

• Hacking Moltbook: The AI Social Network Any Human Can Control
• Does AI already have human-level intelligence? The evidence is clear
• Introducing GPT-5.3-Codex
• Introducing Claude Opus 4.6
• Sam Altman responds to Anthropic’s ‘funny’ Super Bowl ads

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin: https://www.linkedin.com/in/justincollery/
Frank: https://www.linkedin.com/in/frankprendergast/