The Comunicano Sunday Edition for June 29th 2025
Let me be blunt: we are not entering the AI era. We’re already neck-deep in it. And the rules are being rewritten in real time.
When one of the most powerful AI labs decides to shift from Nvidia to Google chips, it’s not just a supply chain tweak. It’s a tectonic signal: the future of AI isn’t locked into one platform, one partner, or one philosophy. It’s diversifying. Fast. And that means opportunity for all of us.
Yes, AI scams are multiplying. Deepfakes are distorting elections. And yes, bad actors are getting smarter. But so are we. Every attack sharpens our defenses. Every fake headline teaches us to question better. AI isn’t just a threat. It’s our next great proving ground.
Look at coding agents. They’re not perfect, but they’re democratizing software creation. They’re turning English into the next programming language. That’s not the death of developers. That’s liberation from gatekeeping. That’s unleashing millions of builders who were never handed the keys.
Look at the travel sector, once a soft target for cybercrime, now leading the charge with AI-powered detection tools. That’s not a weakness. That’s evolution.
And what about the so-called chaos merchants like Cluely? Sure, they game the system. But they also expose its cracks. They remind us that virality without values is a dead end. In that exposure lies our leverage.
The narrative around AI is too timid. Too fearful. We need to reclaim it. AI is not the end of thinking, it’s the beginning of thinking differently. It’s not making us obsolete. It’s making us urgent. We are being dared to level up.
We can build systems that are resilient, education that is personalized, industries that are more secure, and democracies that are informed rather than manipulated. If that sounds impossible, good. That means it’s worth doing.
The future’s here. It’s volatile. It’s wild. And it’s ours to shape—if we choose to lead. Which, after all, is why you read, THE COMUNICANO!!!
Andy Abramson
AI Watch
OpenAI Shifts to Google TPUs (Reuters)—OpenAI has started leasing Google’s Tensor Processing Units (TPUs) via Google Cloud to power ChatGPT and other applications—a significant departure from its heavy reliance on Nvidia GPUs and Microsoft Azure. While not using Google’s top-tier TPUs, the move signals a strategic push to diversify infrastructure, ease rising inference costs, and reduce dependency on a single vendor. This marks the first major adoption of non‑Nvidia hardware and represents a rare collaboration between direct competitors. As OpenAI navigates sky-high demand, Google gains momentum in commercializing its cloud-based AI chips. Read more here
Travel Industry Faces Rising Cyber Threats from AI Powered Scams (PhocusWire)—Booking.com's Chief Security Officer and other experts warn of a surge in AI enhanced phishing and credential stuffing attacks across the travel sector. Criminals now pose as both customers and suppliers, crafting convincing multilingual messages and spoofing legitimate platforms to steal credentials or payments. The travel industry’s fragmented systems and reliance on third party vendors make it especially vulnerable. Hotels, tour operators, and booking platforms are boosting defenses with AI powered detection tools and intensive employee training to counteract the evolving threat landscape. Read more here
Cluely’s Controversial Rise Shows Why ‘Momentum Is the Moat’ (TechCrunch)—Andreessen Horowitz's $15M bet on Cluely—a provocative startup promising tools to “cheat on everything”—has less to do with the product and more with its playbook. Cluely’s founder Roy Lee leans into rage-bait marketing, going viral with AI-powered stunts and controversy. A16z’s Bryan Kim argues that in today’s fast-moving AI world, speed and visibility trump polish. “Momentum is the moat,” Kim says. Cluely launched with hype before even building a real product, but its ability to turn attention into paying users makes it the archetype for post-generative AI startups. Read more here
Hands-On with Seven Coding Agents Reveals AI’s Promise and Pitfalls (Understanding AI)—Timothy B. Lee tested seven agentic coding platforms—from Bolt to Claude Code—and found massive variation. While tools like Bolt struggled with large datasets and Replit failed to render a working site, Claude Code excelled, completing tasks with minimal backtracking and high accuracy. Cursor followed closely, especially when given structured context. Lee concludes that English is fast becoming the new programming language: agents amplify productivity but require precision, planning, and domain expertise. Coding agents won’t replace developers—yet—but they’re redefining how software is created and who can build it. Read more here
Generative AI Emerges as a Global Threat to Elections (Economic Times)—Generative AI is reshaping political discourse—and not in a good way. Free tools are generating deepfakes and voice clones that have polluted democratic elections in over 80 countries, according to international researchers. From Romania’s nullified election to AI-backed far-right surges in Germany and Poland, the technology has enabled the rapid dissemination of misinformation at scale. While some candidates use AI constructively, 69% of its use in elections has been harmful. With governments and social platforms slow to act, and AI tools growing more powerful, experts warn democracy itself is being eroded. Read more here
AI Avatars Reshape Workplace Learning for Gen Z (Fortune)—AI-generated avatars are revolutionizing corporate training, especially among Gen Z employees. Tools like Synthesia and HeyGen now enable companies to produce internal videos with customizable digital presenters at scale, replacing costly video shoots. These avatars offer multilingual voiceovers and brand-consistent appearances, streamlining corporate communications while catering to digital-native preferences. More than 70% of Fortune 100 companies are using such platforms to speed up content delivery and boost engagement in a cost-efficient way. This marks a shift to scalable, on-demand training driven by AI and personalized branding. Read more here
ChatGPT Watch
ChatGPT May Dull Brain Engagement, MIT Study Finds (Washington Post)—An MIT study shows that students using ChatGPT for writing tasks displayed the least brain activity and poorest memory recall. Compared to search engines or working unaided, ChatGPT users were least engaged, struggled to reproduce their work, and produced generic-sounding essays. Researchers used EEG scans to measure neural activity and found that AI use may lead to reduced cognitive effort and long-term memory loss. Experts worry this trend could affect creativity, learning retention, and students’ confidence in their own thinking. Read more here
OpenAI is hopeful GPT‑5 will compete a little more—OpenAI shared at a recent summit in Mexico that GPT‑5 is under active development and poised to outmatch current models. The goal is to better rival offerings like Gemini 2.5 Pro and Claude 4, especially in areas like coding. Although pricing details are still under wraps, it won’t be cheap. A summer launch, possibly July, remains the target, assuming internal benchmarks are cleared. With a focus on seamless feature integration and improved ecosystem performance, OpenAI is aiming for a more competitive and capable AI. Read more here
China Watch
China Closing Algorithm Gap with the U.S. Despite Chip Shortage (SCMP)—Harry Shum, former AI head at Microsoft, stated China is closing in on the U.S. in AI algorithm capabilities despite a persistent chip shortage. Speaking at a summit in Hong Kong, he pointed to startups like DeepSeek, which rival top U.S. models using a fraction of the computational resources. While China’s hardware limitations remain, Shum emphasized software optimization is enabling rapid progress. The implications are clear: algorithmic efficiency may offset hardware dominance in the AI race. Read more here
Meta Watch
Meta’s $100M Bids for AI Geniuses Target ‘The List’ (Wall Street Journal)—Meta CEO Mark Zuckerberg has created “The List,” a confidential roster of elite AI researchers he's actively pursuing with compensation packages reportedly reaching $100 million. As part of Meta’s Superintelligence lab push, Zuckerberg has offered massive deals to poach talent from rivals like OpenAI and Google DeepMind. While not every offer hits nine figures, they often combine salaries, equity, and performance-based incentives. Top-tier researchers now operate like free agents, quietly comparing deals and forming package arrangements. With Meta investing $70B in AI this year, human capital looks like a bargain next to chips and compute. Read more here
Meta and the Myth of $100M AI Signing Bonuses (Spyglass)—A reported $100M signing bonus offered by Meta to lure OpenAI talent triggered a wave of media speculation—fueled by Sam Altman’s podcast comment and later rebutted by Meta CTO Andrew Bosworth. The truth is murkier: while such eye-popping figures are likely part of total compensation packages—blending cash, stock, and performance incentives—they are not singular "sign-on" bonuses. Meta has reportedly made offers exceeding $100M to key hires, including Scale AI’s Alexandr Wang. The drama underscores a frenzied AI talent war where compensation narratives become both strategic tools and internal flashpoints. Read more here
Microsoft Watch
Microsoft to Purge Saved Passwords—Passkeys Are the Future (CNET)—Microsoft will delete all saved passwords in its Authenticator app by July, urging users to migrate credentials to Edge or adopt passkeys. This bold move aims to accelerate the shift away from traditional logins toward more secure, phishing-resistant passkeys tied to devices and biometrics. Users must act quickly to sync their data or risk losing access. With passwords increasingly targeted in breaches and reused across sites, Microsoft’s push reflects a broader industry trend prioritizing user safety and simplicity. Read more here
Reddit Watch
Reddit Fights Back: AI, Lawsuits, and Reddit Answers (CNBC)—As AI chatbots like ChatGPT reshape how users get information, Reddit is fiercely defending its data. The platform has filed a lawsuit against Anthropic, alleging the AI firm scraped subreddit content for commercial use without permission. Reddit asserts its trove of user-generated content is a valuable asset—and it's monetizing it through data licensing deals with OpenAI and Google. In parallel, Reddit launched its own AI tool, Reddit Answers, offering responses rooted in subreddit discussions. With over one million weekly users, it blends human insights with AI convenience. Read more here
Tesla Watch
Tesla just delivered itself to a customer autonomously, Elon Musk confirms (Tesarati)—Tesla hit a self-driving milestone as a Model Y drove itself from the Austin Gigafactory to a customer, without a driver or remote oversight. Elon Musk confirmed the autonomous delivery happened a day early, with the car navigating a 15-mile journey across highways, surface streets, and parking lots, topping out at 72 mph. While Tesla hails this as a world first, rivals like Waymo have already operated driverless rides. Nonetheless, this event showcases Tesla's FSD progress, even as regulators scrutinize recent robotaxi safety issues. A bold step, albeit one still surrounded by operational caution. Read more here
Tesla Robotaxi launch was a scam (AI Supremacy) —AI Supremacy pulls no punches, calling Tesla’s Robotaxi rollout a scam built on hype and flawed vision. The article argues Musk’s promises never aligned with technological reality, spotlighting a decade of overreach. It critiques Tesla’s camera-only autonomy approach and claims media influencers, not journalists, shaped the narrative. The piece questions the safety and ethics of deploying unproven robotaxis, asserting no informed expert would ride in one. For AI Supremacy, the launch was more a matter of theater than a tech milestone, a cautionary tale of hype outpacing substance. Read more here
Streaming Watch
DIRECTV Enhances Genre Packs with Paramount Channels (The Streamable)—DIRECTV has expanded its Genre Pack bundles by integrating nearly 20 Paramount networks such as Nickelodeon, CBS, MTV, Comedy Central, and BET into MySports, MyEntertainment, MyNews, MiEspanol, and launching a new $19.99/month MyKids pack (including Disney Channel, Cartoon Network, Nick Jr., plus ad-supported Disney+). Existing pack prices remain unchanged. Available since June 26, the upgrade enhances value for cord-cutters seeking focused, affordable viewing options. Read more here
Cybercrime Watch
How Global Scammers Exploit U.S. Banks in $44B Fraud Web (ProPublica)—ProPublica reveals how Asian crime syndicates orchestrate pig-butchering scams, siphoning $44 billion annually from victims via U.S. banks. Fraudsters open fake accounts using stolen business identities, such as Boston’s Middlesex Truck, then launder funds through cryptocurrency. Telegram channels facilitate account rentals, and banks like Chase and Bank of America have been implicated for lax oversight. One victim, Kevin, wired $716,000 across multiple institutions, only recovering funds after media exposure. Experts warn weak U.S. regulations and voluntary fraud reporting standards make the banking system vulnerable, with little recourse for victims. Read more here
Travel Brands Grapple with AI-Fueled Cyber Threats (PhocusWire)—The hospitality sector, built on trust, is now a ripe target for cyberattacks accelerated by generative AI. At Phocuswright Europe, Booking.com’s CSO Marnie Wilking warned that scams from AI-crafted phishing emails to credential stuffing have surged since 2022, with some reports citing a 1,000% increase. Attackers use AI to craft multilingual, convincing emails and scripts to exploit reused passwords across services. Nation-state actors now target small vendors to infiltrate larger partners. Wilking emphasized defense via prevention, detection, and education while also harnessing AI to strengthen cybersecurity measures. Read more here
