GPT-5 is here: hype, AGI… or just another upgrade?

Published on August 9, 2025 · Duration 00:44:24

Welcome back, fellow tech enthusiasts and curious minds, to another electrifying blog post from the Bootstrapped for Cash podcast! This week, your hosts Giorgio Azzinnaro and Iván Peña dive deep into the most anticipated AI release in recent history: OpenAI’s GPT-5. Has the era of Artificial General Intelligence finally dawned, or is it just another impressive, albeit polished, update? Get ready for some laughs, a healthy dose of skepticism, and all the juicy details from the AI world.

GPT-5: The AGI We’ve Been Waiting For?

OpenAI dropped GPT-5 with much fanfare, promising a leap towards AGI. But Iván, ever the pragmatist, spent an hour and a half watching the keynote only to conclude it was… well, an hour and a half he won’t get back. The episode kicks off dissecting the “new” GPT-5, now the default model in ChatGPT, merging previous iterations into one hybrid system. Is this for user experience, or a clever cost-saving maneuver by OpenAI to reduce token consumption? The hosts lean towards the latter, given past pleas from OpenAI about going bankrupt from too much politeness!

Key takeaways from the GPT-5 announcement include:

• Hybrid Model & Default Status: GPT-5 is now the single default model, choosing whether to “think” or not on its own. While marketed as a UX improvement, it’s likely a shrewd cost-saving strategy.
• Wider Access: OpenAI claims to democratize access, even for free-tier users, which again, our hosts suspect might be more about managing costs than saving the world.
• Reduced Hallucinations & Speed: Benchmarks were shown, but conspicuously, only against OpenAI’s own previous models. Ivan highlights that in crucial tests like ARC AGI and Humanity’s Last Exam, existing models like Grok 4 actually outperform GPT-5.
• Coding Prowess: OpenAI heavily emphasized GPT-5’s coding capabilities, even inviting the CEO of Cursor. However, early tests by Giorgio found it “the usual sick of fund,” with generic feedback.
• New Features, Old Tricks?: Advanced voice mode, Gmail/Google Calendar access for personal assistance, and “safe completions” (already jailbroken, apparently) were showcased. Yet, many of these features are already available in competitors like Grok 4 and Claude.
• Healthcare Dilemma: The discussion turns serious with GPT-5’s touted use in healthcare. While presented as a tool to help cancer patients understand medical jargon, the hosts raise a red flag about the dangerous precedent of people using AI as a doctor replacement.

In summary, the AGI hype surrounding GPT-5 seems to be just that – hype. The hosts conclude that OpenAI is still “not even close” to achieving AGI, and in many aspects, appears to be playing catch-up with other providers.

Other AI Shenanigans and What’s New in the AI World

The AI landscape is always buzzing, and this episode covers a whirlwind of other intriguing, and sometimes alarming, news:

• ChatGPT & Confidentiality: Sam Altman warns users that there’s no legal confidentiality when using ChatGPT as a therapist or lawyer. So, think twice before spilling your deepest secrets or legal woes!
• Public Chats Backlash: Remember the scare about your private ChatGPT conversations being indexed by Google? OpenAI quickly removed the option to make chats discoverable by search engines after a significant backlash, citing users “accidentally” enabling the feature.
• Anthropic vs. OpenAI: The drama continues as Anthropic cut off OpenAI’s access to its Claude models, apparently because OpenAI was using them to train GPT-5’s coding skills. This mirrors the earlier “Windsurf telenovela,” where Anthropic also cut off Windsurf’s access before its ill-fated acquisition attempts.
• Windsurf Employees’ Woes: Following its acquisition by Cognition, former Windsurf employees are being offered buyouts or, as the alternative, “inhumane working conditions” of 80 hours a week. Ouch!
• Google’s Deep Think & Genie 3: Google released Gemini 2.5 Deep Think, an “agentic system” that spins up multiple agents to work on the same task, aiming to reduce hallucinations by finding a consensus. While the hosts are cautiously optimistic about its potential for reasoning problems, they’re not fully convinced about its hallucination-reducing claims. Plus, Google Deep Think officially clinched a gold medal in the International Mathematical Olympiad, showcasing its mathematical prowess. Not to be outdone, Google also released Genie 3, a model that generates immersive, interactive virtual worlds in real-time. Imagine creating your own GTA 6 with a prompt! This could spell trouble for traditional game developers and Hollywood alike, as users could potentially control plots and endings of movies and games.
• Meta’s Superintelligence Claims: Mark Zuckerberg announced that Meta’s internal models are starting to improve themselves, hinting at “superintelligence”. Our hosts are… skeptical, to say the least. “For me, it’s clear,” Iván says, “it’s just marketing.”
• Microsoft’s AI Impact on Work: A new paper from Microsoft analyzes AI’s impact on various jobs. Unsurprisingly, interpreters and translators are highly affected, but some surprising less-affected roles include “dredge operators” and “bridge and lock tenders”. Go figure!
• Grok’s Spicy Mode: And for the grand finale, xAI’s Grok image and video generator now includes a “spicy mode” for Heavy and iOS users, which many are calling “porn on demand.” This comes from Elon Musk, who has previously warned about declining birth rates – quite the conflicting message, wouldn’t you say?

Bootstrapped for Cash: Product Updates

In podcast news, the team is making strides with LiveMonetize! They’re working towards becoming a PayPal partner to handle more countries and improve their payment system, a notoriously “nightmarish” endeavor. They’ve even established an LLC in Delaware to navigate payment restrictions. While a recent application to Meta was, predictably, rejected, they’re not deterred.

Giorgio and Iván also discuss their LinkedIn Premium trial. While Iván saw one post go “viral” with nearly 9,000 impressions (a personal record!), he’s skeptical if it’s the premium features or just a good idea. They advise that LinkedIn Premium might only be worth it if LinkedIn is your primary B2B distribution channel.

That’s a wrap for this week’s whirlwind tour of the AI universe! Don’t miss the full, unedited, and even funnier discussion by tuning into the latest episode of Bootstrapped for Cash. Leave us a review or a comment – we’d love to hear from you and keep making the podcast better!