An absolutely breathless week in the world of artificial intelligence! If you thought the tech space was moving fast last month, think again, because the past week was a sprint. The biggest news of the week was Anthropic getting the green flag to re-release Fable 5 and Mythos 5 – you can now use Anthropic’s most capable model again.
But that’s not all from this week!
We are steadily moving from simply talking to algorithms to letting them decode our brainwaves, run our code via physical desktop macro pads, and completely reshape corporate technology budgets. It feels like we are living through a sci-fi script written in real-time, where the lines between science fiction, consumer hardware, and global multi-billion-dollar corporate tech contracts have completely dissolved into thin air.
From Washington D.C. to India’s bustling IT corridors, here is everything mind-boggling that went down in AI this week, brought to you by your favourite AI gossipmonger.
Claude Fable 5 is back, and so is Mythos 5
Let’s start with Uncle Dario and his mythological friends. Following stringent security evaluations and regulatory clearances, the US government lifted export restrictions on Anthropic’s high-tier Fable and Mythos models, paving the way for a massive global rollout of Mythos 5. After discussions on AI sovereignty, individuals and enterprises across India are rushing back to integrate Anthropic’s mythological AI models.
Nonetheless, Mythos 5 and Fable 5 are expected to help enterprises and individuals across the world to stress-test critical infrastructure and defend high-stakes corporate networks against sophisticated, AI-generated cyber threats. And not to forget, these models can create anything in hours that humans and existing AI models would have taken months – from video game clones to vibe-coding apps, there’s everything here.
Zuck wants in your brain and your Pocket
No, I ain’t referring to the addiction issues that Zuckerberg is usually accused of all the time. Uncle Mark wants to take on Uncle Elon’s Neuralink and hence, Meta AI dropped Brain2Qwerty v2 this week as an end-to-end AI pipeline that can decode non-invasive brain activity into actual text. Confused? In plain English, this AI model can read your mind without implanting a chip in yoru brain!
Using magnetoencephalography (MEG) to track neural signals while a subject types, the AI achieved a staggering 61% average word accuracy (jumping to 78% for top participants) without requiring a single surgical implant. It’s an incredible breakthrough for medical accessibility, but a slightly terrifying glimpse into a future where your inner thoughts have an auto-correct feature.
Meanwhile, for those who prefer to keep their thoughts in their skulls and in their pockets, Meta also quietly launched Pocket on the App Store and Google Play on June 29. Born from Meta’s stealthy acquisition of the team behind Gizmo, Pocket allows everyday users to “vibe-code” small, interactive games and media experiences entirely out of text prompts. You type what you want to play, Meta builds it, and your friends can remix it instantly.
Uncle Sam’s first hardware device is Codex Micro
Uncle Sam has shown its first physical product and it’s not the one designed by Jony Ive. OpenAI teamed up with boutique keyboard manufacturer Work Louder to create Codex Micro – a compact, physical macro pad designed explicitly for developers. It features mechanical keys and dial shortcuts mapped directly to your IDE. Instead of constantly typing prompts or hunting through drop-down menus, vibe-coders can trigger code generation, debugging workflows, and code summaries with a single physical button press.
This hardware play comes alongside staggering corporate maneuvers. Reports emerged that CEO Sam Altman has proposed giving the US government a massive 5% equity stake in OpenAI, valued at roughly $42.6 billion, to seed a public “Public Wealth Fund” while simultaneously dropping a highly controlled preview of its next-gen flagship suite: GPT-5.6 Sol.
OpenClaw goes mobile
Speaking of on-the-go agents, the OpenClaw Foundation officially dropped native mobile apps for Android and iOS, aiming to put private, self-hosted AI personal assistants right in your pocket. The promise is massive – push-to-talk voice commands and the ability to approve or deny automated agent workflows straight from your phone’s notification shade.
However, the early feedback has been brutal. Early adopters have taken to social media to label the initial rollout a “buggy mess” with a highly unpolished interface. It’s local-first AI, but you might want to wait for patch 1.1 before letting it loose on your emails.
Corporate India’s billion-Dollar deals and “Token Bleed”
Over in India, the financial momentum behind AI is officially exploding into the stratosphere. HCLTech grabbed headlines by locking down a massive $1.14 billion, six-year contract with a Europe-headquartered Fortune Global 50 giant to fully orchestrate their digital workplace using advanced, AI-driven operating models.
But it’s not all blank checks. A quieter, much more financially cutthroat phase of AI adoption took off across Indian boardrooms this week. CFOs are starting to sweat over “Token Bleed.” Generative AI tools consume computing tokens at a rate that has caught corporate budgets completely off guard. Indian firms are rapidly shifting to strict ‘token-to-business-value economics,’ putting caps on employee usage of agentic tools or swapping to cheaper, local architectures to stave off budget exhaustion.
It’s clear we’re no longer just building chatbots. We’re building an entirely new corporate and cultural architecture for the planet, and the skill threshold is rising by the hour.
Until next week, keep your tokens guarded and your agents sandboxed!