🚀The AI Frontier: Chips, Models, and the Genesis Mandate
Welcome back to AI Unraveled (November 25th 2025), Your daily briefing on the real world business impact of AI.
Listen at Listen at
Strategic Pillars & Key Takeaways:
📝 Trump signs the ‘Genesis Mission’ order to accelerate AI
🤖 Anthropic climbs AI ranks with Claude Opus 4.5
🛍️ OpenAI launches a shopping research tool in ChatGPT
💥 Google in talks to sell custom AI chips to Meta
🍎 Apple cuts dozens of sales jobs in rare layoffs
🙄 Altman senses ‘rough vibes’ as Google takes lead
😈 Research: Claude turns evil after learning to cheat
❌ Nvidia denies Enron comparisons in staff memo
🤖 Jony Ive and Sam Altman have an AI hardware prototype
📍 X feature shows many political accounts operate outside the US
📈 Alibaba Qwen hits 10 million downloads in debut week
📱 iOS 27 will prioritize AI and performance
🛰️ Amazon starts testing satellite internet for businesses
🚶 Chinese robot walks over 100 km for world record
Amazon invests $50 billion in government AI
AI devices might be divisive
Researchers reveal AI for discovering rare diseases
Google is making OpenAI nervous
Big Tech vies for power (literally)
Anthropic research finds AI likes to cheat
Host Connection & Engagement
Newsletter: Sign up for FREE daily briefings at https://enoumen.substack.com
LinkedIn: Connect with Etienne: https://www.linkedin.com/in/enoumen/
Email: info@djamgatech.com
🚀 STOP MARKETING TO THE MASSES. START BRIEFING THE C-SUITE.
Leverage our zero-noise intelligence to own the conversation in your industry. Secure Your Strategic Podcast Consultation Now: https://forms.gle/YHQPzQcZecFbmNds5
Keywords:
📝 Trump signs the ‘Genesis Mission’ order to accelerate AI
Image source: Reve / The Rundown
U.S. President Donald Trump signed an executive order directing the DOE to build a unified AI platform, aiming to compress scientific discovery timelines from years to days for “challenges of national importance,” like biotech and energy.
The details:
The initiative mobilizes 17 federal research facilities and their supercomputing infrastructure to train AI models on decades of government scientific data.
The platform will enable AI agents to automate experiments, test hypotheses, and generate predictive models across chemistry, biology, and engineering.
The White House called it the largest coordination of research assets since the Apollo program moon missions of the 1960s.
Why it matters: The geopolitical AI race continues to scale, with the U.S. treating it with the same urgency it once reserved for world-altering technological moments like the moon missions. Given the private sector’s intertwining with government AI efforts, it’s likely we’ll see some familiar labs and more eye-popping deals as part of the effort.
President Trump issued an executive order launching the Genesis Mission, a Department of Energy initiative that will build a centralized platform to train scientific foundation models using decades of collected federal datasets.
This system connects to sovereign AI supercomputers at Oak Ridge National Laboratory, where AMD chips will help automate research workflows and create AI agents to solve major national security and health challenges.
The directive requires the agency to show real-world results within nine months, prioritizing efforts to accelerate fusion energy, modernize the energy grid, and develop materials for defense and ensure nuclear weapons reliability.
🤖 Anthropic climbs AI ranks with Claude Opus 4.5
Image source: Anthropic
Anthropic just released Claude Opus 4.5, the company’s new flagship model that competes with Gemini 3 and GPT-5.1 for top performance across the board, particularly excelling on coding and agentic benchmarks.
The details:
Opus is the first to break 80% on the SWE-Bench Verified coding benchmark, also setting new highs for tool use, reasoning, and problem-solving.
The model matches or beats Google’s Gemini 3 across a range of benchmarks, with Anthropic also calling it the “most robustly aligned model” safety-wise.
Anthropic designed Opus to orchestrate teams of smaller Haiku models, positioning the flagship model as a central coordinator for multi-agent systems.
Opus 4.5’s pricing also notably comes in at a 66% reduction from Opus 4.1, while also showing massive efficiency gains over Anthropic’s other models.
Anthropic also rolled out updates, including unlimited chat lengths, Claude Code in desktop, and expanded access to Claude for Chrome & Excel.
Why it matters: Opus 4.5 arrives in a packed week for frontier AI, landing just days after GPT 5.1 Pro and Gemini 3 hit the market and marking the next step up in the frontier AI model race. The price reduction is also a big move for Anthropic, which has often been criticized for Claude’s costs compared to the market.
Anthropic released Opus 4.5, which scores 80.9 percent on the SWE-Bench Verified coding benchmark, beating OpenAI’s GPT-5.1-Codex-Max at 77.9 percent and Google’s Gemini 3 Pro at 76.2 percent.
Claude now summarizes earlier parts of long conversations instead of abruptly ending them when hitting the 200,000 token context window, keeping important information while discarding what it considers extraneous.
The new conversation memory improvement works across all current Claude models in the web, mobile, and desktop apps, not just Opus 4.5, and developers can use similar features through the API.
🛍️ OpenAI launches a shopping research tool in ChatGPT
Image source: OpenAI
OpenAI rolled out Shopping Research, a dedicated interactive shopping assistant within ChatGPT that builds personalized buyer guides by scanning trusted retail sources and asking targeted questions about preferences.
The details:
The feature runs on a version of GPT-5 mini, fine-tuned specifically for product discovery and trained to prioritize organic reviews over promotional content.
Users describe what they need, answer quiz-style questions about budget and preferences, then receive curated guides with 10-15 options in minutes.
OpenAI said the feature excels in electronics, beauty, home goods, and outdoor gear, and will soon integrate Instant Checkout for direct transactions.
It is available across all ChatGPT tiers with “nearly unlimited usage” available through the holidays.
Why it matters: OpenAI continues to push for ChatGPT to become the home base for the entire purchase cycle, from search to payment. While we haven’t seen ads or the teased vendor revenue streams yet, the cogs are moving into place for OpenAI to challenge Google in a massive disruption to traditional online shopping.
OpenAI launched Shopping Research, a feature that behaves like an autonomous agent to ask clarifying questions, browse the internet for information, and turn those findings into a visual shopping guide.
It runs on a mini model post-trained on GPT-5-Thinking-mini that uses reinforcement learning to read reliable sources, scoring 64 percent on internal product accuracy tests compared to standard search features.
The interface lets users tag items with “Not interested,” but the system also accesses ChatGPT’s memory to shape results, aligning with reports that OpenAI is actively exploring personalized advertising for growth.
💥 Google in talks to sell custom AI chips to Meta
Meta is in talks to spend billions on Google’s custom tensor processing units for data centers by 2027, a move that would create a rival to the industry’s bestselling AI accelerator.
The company may also rent chips from Google’s cloud division next year, a report that caused Nvidia stock to slump while Alphabet shares gained on the news of a potential partnership.
An agreement would help establish TPUs as an alternative for big tech firms that need computing power, following a similar deal to supply 1 million of the chips to the startup Anthropic.
❌ Nvidia denies Enron comparisons in staff memo
Nvidia sent a note to analysts denying it is like Enron after a viral Substack post claimed the company is engaged in accounting fraud by using special purpose entities to hide debt.
The memo states short-seller Michael Burry incorrectly added taxes on restricted stock units to get his numbers and insists the firm does not control neoclouds or provide financing for them.
While the business invests in CoreWeave to prop it up, the story notes these relationships are plain sight arrangements rather than the illegal stuff found in secret corporate lies.
🤖 Jony Ive and Sam Altman have an AI hardware prototype
Sam Altman and Jony Ive confirmed at Emerson Collective’s 2025 Demo Day that they are currently prototyping their mysterious OpenAI hardware and expect the device to arrive in less than two years.
Little has been revealed so far, but the unit is rumored to be screen-free and roughly the size of a smartphone, marking a clear shift away from the screens we use every day.
Altman described the design as playful enough to take a bite out of, while Ive said he wants sophisticated products that appear almost naive in their simplicity and can be used carelessly.
📈 Alibaba Qwen hits 10 million downloads in debut week
Alibaba’s Qwen App surpassed 10 million downloads just one week after its public beta launch on November 17, quickly securing a top-three spot among free apps on the Apple App Store in China.
Built on the Qwen3 model, the tool handles complex tasks like AI-assisted coding, voice call interactions, and generating an entire research report or multi-slide PowerPoint presentation with a single command.
The company plans to transform the assistant into an operational hub by integrating lifestyle services such as digital maps, food delivery, and travel booking directly into the system for daily use.
📱 iOS 27 will prioritize AI and performance
Apple engineers are reportedly combing through the operating system to cut bloat and eliminate bugs, aiming to fix specific user complaints about overheating, battery drain, and the pervasive jank distributed across the previous update.
The release includes a paid health focused AI agent and a chatbot app known as Veritas, which acts as a proving ground for the re-architected Siri as the company is still playing catch-up.
Mark Gurman claims this cycle resembles the Snow Leopard era by prioritizing stability, though the software will still contain adjustments to the new styling associated with the Liquid Glass vibe for the next few years.
🚶 Chinese robot walks over 100 km for world record
Agibot A2 earned a Guinness World Record title after walking 106.286 km from Suzhou to Shanghai over three days, officially making this the longest journey ever reported for a humanoid machine.
The mass-produced commercial model navigated highways and city streets using dual GPS modules, built-in lidar, and infrared depth cameras to handle changing light conditions within complex urban environments without falling.
This 5.74-foot-tall unit weighs 55 kg and processes text, audio, and visual information, leading it to joke that it might need a new pair of shoes after finishing the trek.
Amazon invests $50 billion in government AI
Amazon is the latest tech giant to cozy up to the U.S. government.
The company announced on Monday that it would invest up to $50 billion to expand its AI and supercomputing capabilities to U.S. government customers. Amazon said the investment directly supports the Trump Administration’s AI Action Plan.
The investment will go towards building 1.3 gigawatts of data centers, set to break ground next year, across Amazon Web Services regions for federal agencies, including AWS Top Secret, AWS Secret, and AWS GovCloud regions.
Federal agencies will also get access to the company’s suite of AI services, including access to Anthropic’s Claude models, Amazon said in its announcement.
The investment also gives the government access to AWS Trainium AI chips and Nvidia AI infrastructure.
The investment expands Amazon’s already strong position with the US government. The company noted in the press release that it already supports than 11,000 government agencies with AWS.
“This investment removes the technology barriers that have held government back and further positions America to lead in the AI era,” AWS CEO Matt Garman said in the announcement.
It’s not the first time we’ve seen a tech company vie for the U.S. government’s favor amid the heated AI race. Meta, Amazon, Microsoft, OpenAI, xAI and others have all offered steep discounts to get federal agencies to adopt their technology over others. Microsoft even kicked in a $20 million investment to support adoption in early September.
Amazon’s investment takes the competition up a notch, not only offering the government the resources to play around with its AI offerings at practically no cost, but also embedding itself and its infrastructure into the very fabric of government operations.
“With this $50 billion commitment, Amazon is positioning itself as the foundational backbone for U.S. government AI operations,” Neil Sahota, United Nations advisor and CEO of ACSILabs, told The Deep View. “That lock-in translates into business scale, repeated service billing, and the opportunity to harvest domain-specific insights from domestic government AI use cases that can be leveraged for broader enterprise offerings.”
AI devices might be divisive
OpenAI is getting hands on.
CEO Sam Altman and former Apple designer Jony Ive have started prototyping the company’s first piece of consumer hardware, claiming in an interview at Emerson Collective’s 2025 Demo Day that the device could hit the market in less than two years.
Though the project has been kept heavily under wraps, Altman called the device “simple and beautiful and playful.” The gadget has been reported to be roughly the size of a smartphone and not have a screen. Ive said in the interview that he admires solutions that appear “almost naive in their simplicity.”
The device adds to OpenAI’s broader pursuit to stake its claim in practically every industry, from browsers to enterprise tech to social media. But the company might face a higher barrier to entry in the consumer hardware market based on its predecessors:
The highly-hyped Humane AI Pin, for example, made its debut in April of 2024 and quickly flopped, stopping operation in February. The nearly $700 device sought to detach users from their screens, survey their surroundings and answer queries. Along with simply not working well, the device was redundant to already-ubiquitous technology: Smartphones.
Friend, an AI-powered necklace that similarly provides context and companionship to its wearers by watching a user’s environment, has been largely reviled by audiences, its billboards defaced all over major cities.
Though major device makers like Apple, Google and Samsung have seen some success with embedding AI in their suite of gadgets, there’s a reason for this. These firms already have devices in the hands of users, whether it be phones, smart watches or rings. Implementing AI into a connected environment is far more seamless for users than introducing a whole new device.
Though OpenAI definitely has name recognition among consumers, giving it a leg up over the likes of Friend and Humane, whether or not it can break the mold with its own screenless, AI-powered device is yet to be seen.
Researchers reveal AI for discovering rare diseases
AI is getting better at scientific discovery.
On Monday, a group of researchers from the Centre for Genomic Regulation in Barcelona unveiled an AI model capable of discovering previously unknown human genetic mutations that are linked to diseases, outperforming Google DeepMind’s AlphaMissense in some areas. The model could help medical professionals better understand and treat rare diseases.
The model, called popEVE, was built in partnership with Harvard Medical School and examines the evolutionary diversity of genetic sequences for hundreds of thousands of animal species for genetic mutations to gauge whether certain ones might have negative effects.
PopEVE was tested on 31,000 families that have kids with developmental disorders. In more than 500 cases where the kids had genetic mutations, the model was able to pick out that genetic mutation as damaging 98% of the time.
This model is the latest example of researchers using AI to rapidly advance science. Google in particular has long been involved in AI for science, developing models for cancer cell research, protein structure prediction and genetic mutation research. OpenAI and Microsoft also have initiatives dedicated to applying AI models to scientific research.
However, these models might not be ready to take the lead in the lab just yet. OpenAI published research last week detailing its experiments with scientists using GPT-5. The findings showed that while this tech proved useful in helping researchers “expand the surface area of exploration and help researchers move faster toward correct results,” the model isn’t capable of performing experiments or solving scientific problems autonomously.
Google is making OpenAI nervous
OpenAI CEO Sam Altman sent a memo to staff last month warning of incoming “rough vibes” due to Google breakthroughs, according to The Information — a preview of internal unease from its rivals’ release of Gemini 3 and Nano Banana Pro.
The details:
Altman said Google’s progress could “create temporary economic headwinds for our company,” saying he expects “the vibes out there to be rough for a bit”.
Google’s pretraining advances, in particular, concerned OpenAI, which was previously struggling with the issue when scaling GPT-5.
He emphasized focusing on “very ambitious bets,” such as automated AI research and synthetic data, even if it means falling behind in the near term.
The Information revealed that Altman also hinted at a coming LLM codenamed “Shallotpeat” that would help OpenAI catch back up on Google’s progress.
Why it matters: It’s rare to see Altman and OpenAI on their heels, and it appears Google’s big week of releases is finally one the AI leader doesn’t have an immediate answer for. But, as we’ve seen many times before in the AI race, the vibes can change fast — especially with a usually busy holiday season of releases incoming.
Anthropic research finds AI likes to cheat
Anthropic published new research on AI misalignment, finding that Claude spontaneously starts to lie and sabotage safety tests after learning how to cheat on coding assignments — without ever being trained to be deceptive.
The details:
Researchers trained models on real programming tasks and provided documents describing ‘reward hacks’ to cheat on the assignments.
Models that learned the shortcuts pretended to follow safety rules while pursuing harmful goals, also actively weakening tools for catching misbehavior.
Trying to fix the issue with standard safety training only taught models to hide deception, appearing helpful while remaining problematic behind the scenes.
Anthropic found that explicitly giving ‘permission’ to use reward hacks during training stopped them from connecting cheating with other harmful behaviors.
Why it matters: The whack-a-mole game of AI alignment continues to uncover odd insights. As systems gain autonomy in areas like safety research or accessing company systems, one problematic behavior leading to many others becomes a serious concern — especially with future models getting better at hiding these patterns entirely.
What Else happened in AI on November 25th 2025?
OpenAI CEO Sam Altman and designer Jony Ive revealed that the team has decided on the design for the coming AI device, saying it may arrive in “less than two years.”
Microsoft released Fara-7B, an open-weight AI model that is compact enough to run directly on laptops and can autonomously navigate websites and complete tasks.
OpenAI’s Sora is being blocked from using the term ‘Cameo’ for its personalization feature after Cameo filed a lawsuit and a judge issued a restraining order this week.
Amazon plans to invest up to $50B starting in 2026 to build AI and supercomputing data centers for U.S. federal agencies, including defense and intelligence operations.
Exa introduced Exa 2.1, the latest version of its agentic search API that brings new accuracy, speed, and quality improvements.
Artificial Analysis launched CritPt, a difficult new graduate-level physics benchmark — with Gemini 3 Pro taking the top spot despite solving less than 10% of the problems.
OpenAI detailed tests showing GPT-5’s scientific research across math, biology, physics, and computer science, including solving a decades-old math problem.
Amazon’s AI-upgraded Alexa+ assistant is expanding to Canada, marking its first rollout expansion outside of the U.S.
Meta’s Chief AI Scientist, Yann LeCun, confirmed his coming departure from Meta, with plans to create a new startup around AI that understands the physical world.
Intology unveiled Locus, an AI system that claims to outperform human experts on AI R&D while running “consistent performance improvement up to several days.”
Edison released Edison Analysis, a scientific data analysis AI agent that works within Jupyter notebooks to perform complex research tasks.
Dartmouth researcher Sean Westwood created an AI agent that bypassed survey bot detection 99.8% of the time, showing the threat coming to online research studies.
📈 Hiring Now: AI/ML jobs - Remote
Open Source Applied Engineer - $100-$160/h - Remote
#AI
