The Comunicano for Thursday June 19th 2025
Takeaway Thursday
Let’s talk about the rhythm of our digital lives, because the beat is changing, and not always in tune. At MIT, researchers are sounding the alarm: when we let ChatGPT do the writing, our brains take the backseat. A recent study found that relying too much on AI dulls our memory, our creativity, even our decision-making. It’s not just what we write, it’s how we think that’s changing.
Across the web, the same story plays out. Generative AI is everywhere. Search engines, chatbots, social feeds. But like a band that can’t keep time, it stumbles. Hallucinations, outages, flawed facts. We’re living in what some now call the perpetual beta era—where nothing’s quite finished, and trust is the casualty.
And yet, innovation doesn’t rest. In China, a startup named MiniMax just built an AI model with GPT-4-like skills, at just a fraction of the cost. It's a reminder: this isn't just Silicon Valley’s song anymore. The chorus is global.
Then there's Midjourney’s new video tool. It animates images into short, vivid clips—creative, yes. But like a catchy chorus with thin lyrics, its novelty wears off quickly.
Amazon’s CEO says AI is reshaping everything, from your Alexa to your shopping cart. Over a thousand AI projects are in motion, many replacing human hands. It’s progress—but not without consequence.
Meanwhile, Telegram’s founder Pavel Durov is standing firm against government overreach, defending privacy like it’s a final verse. And Meta? They’re offering $100 million signing bonuses to poach OpenAI’s best minds, while rolling out passkeys that promise easier, safer logins.
But here’s the refrain that matters most: AI is powerful, no doubt. But it’s no substitute for independent thought. For creativity. For judgment. So let’s use these tools, but not lose ourselves in them.
Because in this beta world, thinking critically might just be the most revolutionary act of all.
Stay sharp, stay curious. And, keep reading, THE COMUNICANO!!!
Andy Abramson
AI Watch
ChatGPT May Be Eroding Critical Thinking Skills—Study Finds (Time.com)—A new MIT Media Lab study with 54 participants aged 18–39 compares essay writing using ChatGPT, Google, or no assistance. EEG readings revealed ChatGPT users had the lowest neural engagement—poorer executive control, memory, creativity—and produced formulaic essays over time. Those writing unaided showed stronger brain connectivity and greater satisfaction. Interestingly, when later allowed to use ChatGPT, this group maintained better brain engagement, suggesting balanced AI use may benefit learning. Though the study is small and unpublished, it raises critical questions about how overreliance on AI could dull independent thinking.
Read more here
The Entire Internet Is Reverting to Beta (The Atlantic)—Generative AI is now embedded in nearly every corner of the web—from search engines and chatbots to social media and shopping—yet remains fundamentally unreliable. The Atlantic warns that while tools like ChatGPT and Google's AI overviews are powerful, they frequently hallucinate or display inaccuracies, causing dramatic outages and subtle errors. These janky technologies have led the internet into a perpetual beta state, eroding trust and warning that human critical thinking is at risk. Relying on flawed AI for high-stakes decisions—like healthcare—could have serious consequences.
Read more here
China’s MiniMax Unveils M1: GPT-4 Performance at 0.5% the Cost (Computerworld)—Chinese AI startup MiniMax just threw a haymaker into the LLM cost wars. Its new model, M1, claims reasoning performance rivaling GPT-4 but trained at just $535,000—a fraction of the estimated $100 million OpenAI spent. It leverages Mixture-of-Experts architecture and lightning-fast attention, cutting inference costs and outperforming DeepSeek-R1 on long-context tasks. With 512 H800 GPUs and a lean three-week training window, MiniMax is proving that you don’t need Silicon Valley's budget to disrupt the frontier. Read more here
MidJourney Watch
Midjourney Rolls Out First Video Model V1 (Midjourney Updates)—Midjourney has introduced its first AI-driven video tool, V1, which animates still images into short videos—up to 20 seconds—via a real-time, open-world vision model. The model, accessible via Discord and the web, produces four five-second clips per render with optional extensions. CEO David Holz positions V1 as a stepping-stone toward full 3D and interactive AI worlds. Despite a free rollout, video generation costs eight times more than image output and comes amid legal challenges from Disney and Universal over training data. Read more here
Midjourney Video Hits a Snag (Tom’s Guide)—Midjourney V1 is available on the $10/month tier, letting users animate images with customizable motion intensity and extension options. But unlike competitors such as OpenAI’s Sora, it works only from existing images—not text prompts—and offers shorter clips, putting it behind others in flexibility. Users praise its creativity but note motion novelty wears off quickly. With per-clip costs high and limited input options, V1 is seen as promising yet still early-stage. Read more here
Emergent Optimizer Capabilities in Open‑Weight Language Models (arXiv.org)—A new paper on arXiv reveals open-weight LLMs are beginning to match the optimizer behaviors seen in proprietary models. The study finds these models not only perform complex tasks but also internally develop problem-solving pathways akin to gradient-based optimization. This aligns with theories suggesting that LLMs "learn to think" in ways similar to how neural networks are trained. It’s a significant signal that open-weight architectures may close the capability gap with closed-source giants like GPT-4. Read more here
Amazon Watch
Amazon CEO Andy Jassy on Generative AI's Role in Reshaping the Company (Amazon Newsroom)—Amazon CEO Andy Jassy outlines how generative AI is transforming every facet of the company. From Alexa’s AI-driven assistant to smarter shopping tools like “Buy with Me” and enhanced inventory management, over 1,000 generative AI efforts are now active across the enterprise. Jassy doesn’t mince words: this revolution will reshape the workforce, with automation phasing out certain roles, even as AI talent becomes more critical. The message is clear—AI is Amazon’s new operating system. Read more here
Social Watch
US Resumes Student Visa Interviews with Stricter Social-Media Checks (Washington Post)—On June 18, 2025, the State Department restarted F, M, and J visa interviews with new social-media screening, requiring applicants to set profiles public so consular officers can vet posts for hostility toward US values, support for terrorism, or antisemitism. This move reverses a May pause aimed at expanding vetting and draws backlash over privacy and political discrimination concerns. Critics warn the policy may deter international students, especially from elite universities, and add burdens to embassies. Read more here
Telegram Watch
Pavel Durov on His Arrest and Telegram’s Global Challenges (Le Point)—Telegram founder Pavel Durov discusses his recent arrest in France, the Kremlin’s surveillance efforts, tensions with the Macron administration, and pressure from the FBI. He asserts his detention stemmed from Telegram's refusal to provide Russian authorities with encryption keys and user data. Durov accuses Western agencies of imposing on European sovereignty while praising Telegram’s decentralized, secure model. He warns that central bank digital currencies and AI-enhanced government surveillance pose existential threats to privacy. He reaffirms his commitment to resisting state overreach, positioning Telegram as a global fortress for free speech and encrypted communication. Read more here
Meta Watch
Introducing Passkeys on Facebook for an Easier Sign-In (Meta Newsroom)
Meta has launched passkey support on Facebook’s Android and iOS apps as of June 18, 2025. Users can opt for biometric or PIN-based login, bypassing traditional passwords prone to phishing and reuse. Setup and management are handled in the Accounts Center, and integration with Messenger is coming soon. Authentication aligns with industry standards like WebAuthn and supports secure autofill in Meta Pay. Though passwords and 2-factor remain available, Meta joins major platforms like Google, Apple, and WhatsApp in advancing password-free login. Read more here
Facebook Adds Passkey Support to Counter Phishing and Password Leaks (Digit.in)
—Facebook now enables passkey authentication on mobile platforms, replacing passwords with biometrics or PINs. This change bolsters defenses against phishing and credential theft. The feature will expand to Messenger soon. Though traditional login methods remain, passkeys offer easier and more secure access. Read more here
OpenAI Watch
Sam Altman: Meta Offered $100 Million Bonuses to Recruit OpenAI Staff (Reuters)—Sam Altman revealed on his brother’s Uncapped podcast that Meta approached top OpenAI employees with signing bonuses reportedly up to $100 million. Despite this, none of OpenAI’s key staff have accepted the offers. Altman described this recruitment push as part of Meta's bid to escalate its AI capabilities, but emphasized that building a strong innovation culture requires more than high pay. Read more here
Dive into “The OpenAI Files” — Unraveling the Controversial Shift (Opentools.ai)
“The OpenAI Files,” a 10,000-word exposé by Tyler Johnston and nonprofit tech watchdogs, compiles public documents charting OpenAI's transformation from nonprofit to for-profit. It includes detailed governance charts and raises issues around executive compensation, leadership integrity, and board conflicts of interest. Published independently of OpenAI or its rivals, the report seeks to illuminate how the organization’s structure and culture have shifted—leaving readers to form their own conclusions. Read more here
The OpenAI Files Will Help You Understand How Sam Altman’s Company Works (The Verge)—A companion piece on The Verge highlights the release of “The OpenAI Files,” noting its ambition to demystify OpenAI’s internal evolution, including governance and cultural concerns. The Verge calls it a valuable resource for anyone tracking how the company operates behind the scenes. Read more here
Repository Access (OpenAIFiles.org)—The OpenAI Files Repository (openaifiles.org)
The complete “OpenAI Files” report is hosted on its dedicated website, openaifiles.org. This public repository offers full access to the documents, charts, and source material referenced in the analysis. Read more here
ScaleAI Watch
Scale Responds to Meta Deal (Scale.com blog)—On June 18, Scale AI addressed concerns following Meta’s $14.3 billion investment. The company affirmed its independence, stating Meta won’t access Scale’s internal systems or customer data. Interim CEO Jason Droege emphasized continued operations and commitment to confidentiality, integrity, and “agentic” AI. Scale reiterated its business diversification and dedication to safeguarding partners alongside building data-driven AI products. Read more here
Google, OpenAI, xAI Pause Scale AI Work After Meta Investment (Business Insider)—Following Meta’s $14.3 billion deal for a 49 percent stake in Scale AI, key clients—including Google, OpenAI, and Elon Musk’s xAI—halted or paused projects with Scale, citing data confidentiality concerns. Contractors report sudden pauses, with one investor divesting from Scale. Meta and Scale maintain the partnership won’t compromise impartiality, but the move has triggered a significant shift, prompting competitors to eye Scale’s clients. Read more here
Apple Watch
Craig Federighi on iPadOS, Multitasking, and the iPad’s Essence (MacStories)—
Craig Federighi, Apple’s senior VP of software engineering, explains in a deep dive with Federico Viticci that iPadOS 26 brings productivity closer to Mac-like multitasking while maintaining clear device distinctions. He shares why Apple isn’t merging iPadOS and macOS, using the “spork” analogy to emphasize that combining them would sacrifice strengths. Federighi outlines the evolution of multitasking tools like Split View, Stage Manager, and the latest windowing capability. For power users seeking both portability and functionality, he suggests owning both devices. Read more here
Marketing Watch
Taylor Swift’s Marketing Playbook for Modern Brands (MarketingProfs)—Taylor Swift isn’t just topping charts—she’s writing the rules for modern brand strategy. From surprise album drops to social media storytelling and fandom-powered buzz, Swift’s success comes from blending emotional authenticity with calculated scarcity. Her campaigns are built for virality, but grounded in connection. Marketers take note: her blend of digital intimacy, aesthetic evolution, and community-first messaging is a masterclass in engagement. Read more here
