GPT Atlas: OpenAI’s Browser and the “Pivot” Question
OpenAI has just released Atlas, a web browser powered by ChatGPT.
Atlas integrates ChatGPT directly into the browsing experience with built-in memory, page-aware assistance, and agentic task execution. Users can summarize, plan, and automate tasks without ever leaving the page. Features include “Agent Mode” (to perform actions like filling forms or opening tabs) and privacy controls such as incognito mode, page visibility toggles, and memory management.
But beyond features, this move feels strategic.
OpenAI’s mission was once about AGI (Artificial General Intelligence), now, it’s building browsers, chip partnerships, and infrastructure products. It’s hard not to see this as a pivot toward becoming a full-stack tech company, directly challenging Google on its home turf: search and web access.
Is this pragmatic expansion, or a quiet admission that AGI isn’t close enough yet?
Time will tell, but the parallel with Uber’s shift from autonomous cars to logistics is hard to ignore.
Atlas is available on macOS (Windows, iOS, Android coming soon), and works with all ChatGPT plans.
Google’s Willow Chip and the Quantum Leap
Google just proved something historic: a quantum computer can run a verifiable algorithm faster than any supercomputer, 13,000× faster, to be precise.
Their Willow chip successfully ran the Quantum Echoes algorithm, demonstrating what researchers call a verifiable quantum advantage. In plain terms: results can now be reproduced and verified, which is a huge step toward practical quantum applications.
The implications are massive : From simulating molecules to redefining cybersecurity (where today’s encryption could be cracked in seconds), quantum computing may soon move from theoretical physics to real-world disruption.
Quantum computing has been promising “revolutionary change” for decades. This time, it starts to feel more and more real.
Sora Video App: The Deepfake Dilemma
Sora, OpenAI’s video app, hit 1 million users in just five days, faster than ChatGPT’s own debut.
The app lets anyone generate high-quality videos from text, but it’s already sparking a copyright and ethics storm.
Users have been creating ultra-realistic videos of people who never consented, including deceased celebrities. Others are even using tools to remove Sora’s watermark, erasing signs of AI generation.
OpenAI says it’s working on a revenue-sharing system for creators who license their likeness or IP to the platform, but details are thin.
Meanwhile, legal experts warn this could reshape how copyright and personality rights are enforced.
The race between creativity and regulation is (still) on, and the gap is widening.
Sora’s success shows how much people want accessible video creation tools. But without guardrails, it’s also a glimpse into a deepfake-driven internet.
If you want to learn more about DeepFakes, refer to this guide.
Why LLMs Still Hallucinate
Hallucinations, the confident but false statements made by AI models, remain the Achilles’ heel of large language models.
A recent OpenAI study (2025) argues that hallucinations aren’t just bugs, they’re inevitable. Even with "perfect" training data, the statistical nature of language modeling makes some errors unavoidable.
In short: accuracy will never reach 100%.
Why? Because the training objectives reward models for guessing rather than staying silent when uncertain.
Researchers suggest rethinking benchmarks to reward honest uncertainty rather than overconfident answers.
Until then, LLMs will keep “hallucinating with confidence,” and users must stay vigilant
Albania Appoints the World’s First AI Minister
In a world-first, Albania has appointed an AI-powered minister, named Diella (which derives from the word Sun in Albanian), to oversee public procurement.
Built into the country’s e-Albania digital platform, Diella is tasked to evaluate tenders and ensure transparency in public spending.
Prime Minister Edi Rama calls her “the servant of public procurement,” aiming for 100% corruption-free tenders.
Beyond the symbolism, this is a bold experiment in governance automation.
Whether Diella can resist political “hacking” or bias will be a story to watch.
