In partnership with
In Today’s Issue:
🚨 Geoffrey Hinton bets Google will win the AI race
💡 Poetiq smashes the ARC-AGI-2 ceiling
🍎 Apple's C-suite faces its biggest leadership shake-up since 2011
💥 OpenAI is fast-tracking the GPT-5.2 launch to regain the performance lead
⚔️ Google is breaking its rules by offering its powerful TPUs outside of Google Cloud
✨ And more AI goodness…
Dear Readers,
What if this is the week the AI race actually changes shape rather than just posting another shiny benchmark? OpenAI is reportedly fast-tracking GPT-5.2 as a full-blown “code red” answer to Gemini 3, betting that raw speed, stability, and better reasoning will matter more than flashy demos - and that’s our starting point today. From there, we zoom out: Geoffrey Hinton is openly backing Google to overtake OpenAI, Poetiq just smashed through the ARC-AGI-2 ceiling with a new world record, and Apple is quietly tearing up its org chart as key AI and design leaders head for the exits. Meanwhile, in the background, the really big money war is brewing over chips, with Google’s TPUs lining up as the first serious crack in Nvidia’s armor, plus a video with Anthropic’s resident philosopher, fresh rumors, and a few hand-picked prompts to play with it all - so if you want to understand where the next AI power shift might come from, keep reading.
All the best,
🚨 Hinton Bets Big On Google
Google’s AI godfather Geoffrey Hinton says Google is finally “beginning to overtake” OpenAI, pointing to the momentum of Gemini 3 and the Nano Banana Pro image model, which many in tech now see as surpassing GPT-5 in key capabilities. He argues Google’s end-to-end stack, its own chips, massive data centers, and deep research bench, gives it a structural edge, and even predicts “Google will win” the current AI race. Behind the scenes, Google is doubling down on this bet: it’s exploring a billion-dollar AI chip deal with Meta and funding long-term research with a 10 million CAD donation (matched by the University of Toronto) to establish the Hinton Chair in AI, even as Hinton continues to warn about AI’s existential and labor risks following his 2024 Nobel Prize in Physics win
🫨 Poetiq Smashes ARC-AGI-2 Ceiling
Poetiq has set a new world record on the ARC-AGI-2 benchmark with 54% accuracy on the semi-private test set, becoming the first system to break the 50% barrier. It solved a majority of tasks while slashing the cost to $30.57 per problem, less than half of Gemini 3 Deep Think’s previous SOTA (45% at $77.16). Instead of training yet another giant model, Poetiq builds a meta-system on top of existing LLMs (like Gemini 3 and GPT-5.1), using learned test-time reasoning to intelligently orchestrate calls across models and redraw the cost–performance Pareto frontier.
😿 Apple Rewires C-Suite Amid Exodus
Apple is going through its biggest leadership shake-up since Steve Jobs died in 2011, with over half a dozen senior executives leaving. This includes the heads of AI (John Giannandrea), design (Alan Dye, poached by Meta), legal (Kate Adams), and environment/policy (Lisa Jackson), all direct reports to Tim Cook. The exits span every strategic pillar from chips and AI to design and regulation. While they signal internal tension and an end to an era of extreme stability, they also force Apple to rebuild its top bench for the next wave of AI-first products, making this a high-stakes reset moment rather than just a slow drift.
Anthropic’s philosopher answers your questions
🚨 GPT-5.2 is going to be released on tuesday
The Takeaway
👉 OpenAI is fast-tracking GPT-5.2 as an explicit “code red” response to Google’s Gemini 3 momentum.
👉 The company is pausing side projects to focus on ChatGPT’s speed, reliability, and reasoning quality at the model core.
👉 For developers and companies, GPT-5.2 will likely change real-world behavior through lower latency and more stable outputs, not headline-grabbing tricks.
👉 The OpenAI–Google rivalry is shifting into a continuous upgrade war, where staying competitive means re-benchmarking and adapting on a near-constant basis.
OpenAI just turned a “code red” into a launch plan: GPT-5.2 is now expected as early as Tuesday, December 9, pulled forward from a later December release after Google’s Gemini 3 stole the leaderboard spotlight. The update is pitched less as a shiny new toy and more as a hard refocus on speed, reliability, and smarter personalization for ChatGPT, with internal tests reportedly putting GPT-5.2 back ahead of Gemini 3 on key benchmarks.
Under the hood, GPT-5.2 is a rapid-fire upgrade on top of GPT-5.1, arriving just weeks after that “minor” release, and only a few months after GPT-5 itself, signaling a shift to an almost continuous model rollout rhythm. To make room, OpenAI has reportedly paused or deprioritized other projects like agents and ad products so teams can tune the core model’s reasoning, latency, and stability. In practice, this means developers and companies may soon see faster responses, fewer weird edge-case failures, and models that can adapt more tightly to their own data and workflows.
The question is whether GPT-5.2 can replace the current AI King Gemini 3.0 pro.
Why it matters: GPT-5.2 is a clear signal that frontier AI is now a weekly arms race, not a yearly one, and anyone building on these systems needs infrastructure, processes, and governance that can absorb constant upgrades.
Sources:
🔗 https://www.theverge.com/report/838857/openai-gpt-5-2-release-date-code-red-google-response
🔗 https://www.tipranks.com/news/openai-speeds-up-gpt-5-2-launch-as-googles-gemini-draws-fast-user-growth?utm_source=chatgpt.com
Turn AI Into Extra Income
You don’t need to be a coder to make AI work for you. Subscribe to Mindstream and get 200+ proven ideas showing how real people are using ChatGPT, Midjourney, and other tools to earn on the side.
From small wins to full-on ventures, this guide helps you turn AI skills into real results, without the overwhelm.
🤑 AI Chips, Real Money Wars
A seismic shift in the hardware landscape is incoming because the AI chip monopoly might finally be cracking! Google is breaking its own rules by planning to offer its powerful Tensor Processing Units (TPUs) outside of its proprietary cloud ecosystem, creating a direct challenge to Nvidia’s long-standing dominance.
Until now, if you wanted Google’s specialized chips, you had to use Google Cloud. But now, they are looking to sell or lease these chips for external data centers. While Nvidia’s GPUs are like powerful Swiss Army knives capable of handling diverse tasks, Google’s TPUs are highly specialized efficiency beasts, custom-built to crush the specific matrix multiplication math that powers deep learning.
Industry titans like Apple are already leveraging TPUs, and now Meta is seriously evaluating them to diversify their compute arsenal. Crucially, financial services and tech firms are demanding these chips for their own private data centers to ensure maximum security, a market segment Nvidia currently dominates. This strategy does more than just generate new revenue to chase Nvidia’s enviable 63% operating margins; it creates massive leverage for buyers. Even if companies don't switch completely, the mere existence of a viable, scalable alternative forces Nvidia to rethink its pricing power, finally putting the squeeze on the undisputed king of AI hardware!
Join the 1% who won’t sleep on the future.
Mindstream is the human-first AI newsletter that prioritizes clarity over complexity and progress over perfection—so you can make smarter moves, lead calmer conversations, and keep your team a step ahead, today.
