Editor’s Note
You may have noticed that there was no newsletter last week. That was intentional.
Taking just a day away from AI is enough to see nearly 3 to 5 new releases or products hit the market. Sometimes even a major release from OpenAI, Anthropic, Gemini, or others. But in a week? Without starting an entirely new newsletter, we saw Anthropic releasing Claude Opus 4.7, Mythos from Anthropic being announced, OpenAI’s competitor to Mythos being announced, GPT-Rosalind releasing for STEM, a 40% delay in data centres, Gemini 3.1 Flash releasing and a lot more. Including, what I personally think is one of the most important releases of last week, an updated report from Stanford University around AI adoption in companies and society.
And this is just the major bits of news. Anything below this such as Meta making an AI version of Mark Zuckerberg has been quietly sideswept. What I am trying to say is, in the week that this newsletter didn’t happen, a lot did.
So when I hear from companies that they want to let their CIO handle this or a person with existing responsibilites, I immediately raise an eyebrow and feel a sense of disbelief. It’s a task on its own for one person to stay on top of everything going on, even with AI. It’s next to impossible for a CIO or someone in a high stress position to keep ontop of it and manage their responsibilities
The point of this newsletter is to help you stay informed of what’s going on in the world of AI and Tech. To keep it easy. Which is more important now for high pressure roles than ever before. Especially if you have specialised knowledge.
What happened last week in AI?
Sam Altman closed the largest funding round in history — $122 billion at an $852 billion valuation. The same guy who begged Congress for regulation in 2023 just proved the best regulation is "regulate everyone else while I stack cash." Classic Altman speedrun.
Meta flipped the script and went closed-source with the "Avocado" successor to Llama. After years of handing out weights like free samples, Zuck decided open-source generosity while laying off thousands wasn't the long term vibe. Farm closed for business.
Chinese labs kept cooking with Qwen3.5-Omni. A true omni-modal beast that chews through 10+ hours of audio-visual data across 113 languages. Meanwhile Arm dropped its first CPU built specifically for agentic AI. The chip arms race just got multilingual and multi-front.
Supply chain chaos hit hard: Claude source leaks, Axios npm attacks, LiteLLM vulnerabilities, and even Cisco code getting swiped. Nicholas Carlini warned LLMs could find zero-days... and the week proved the models aren't the only ones exploiting bugs.
Mistral raised $830M in debt to fuel more Nvidia data centers in Europe. No jokes here. Just wish my name was data center.
Enterprise kept going agentic: Shopify wiring millions of merchants into AI agents, Zoom pushing agentic meeting tools, and multi-model Copilot workflows so your apps can now debate themselves before giving you answers. Your middle manager's job description just got shorter.
Trump administration dropped a National AI Legislative Framework to boost U.S. competitiveness. Because when you're already racing China on chips and models, the best move is making sure the rules don't slow down the home team.
Bonus: Mythos-level leaks rumored as step-changes, Stanford warning everyone not to blindly trust AI advice right as white-collar cuts accelerate, and the usual quiet reality checks that the demo era is ending while gross margins become the new benchmark.
And, this is only what happened this week.
What action can you take?
OpenAI raised a record $122B at ~$852B valuation while killing consumer experiments like Sora to focus on agents/superapps. At the same time, Alibaba dropped Qwen3.5-Omni. A native omni-modal model handling 10+ hours of audio/video, 113 languages, with strong real-time capabilities.
In instances like this, you should consider building a "model router with attitude". An internal orchestration layer that dynamically routes tasks to the best (and cheapest/most sovereign) model based on modality, language, sensitivity, and cost. Treat OpenAI as the expensive but reliable enterprise default, Qwen as the multilingual/long-context wildcard, and smaller open models for everything else.Deeper twist: Since Qwen excels at long-context multimodal (meetings, lectures, surveillance-style footage), use it to create "corporate memory agents" that ingest weeks of internal calls/emails/videos and answer questions with near-perfect recall. Combine with the Claude leak insights for self-correcting loops.
The endgame? Your company develops a de facto internal AGI-lite that knows more about your operations than any single vendor model ever could. All while keeping the sensitive data inside your walls.
Prompt of the Week
Guess who’s been rewatching James Bond in light of the upcoming game? Me. I used Google Nano Banana Pro for this image. Attach a selfie or a picture of yourself and the location image. Use this prompt to get the candid effect. Change details in the prompt to suit your taste.
Hyper-realistic street photography shot of a candid young man leaning casually against the front bumper of James Bond's Aston Martin DB5. The man wears astylish all-black outfit consisting of a tailored black blazer, an unbuttoned black shirt, black trousers, black leather gloves and dark sunglasses. The background is the courtyard of The OWO hotel (as attached), Soft, natural overcast lighting, high definition, urban streetwear aesthetic, 8k resolution.
🏆 Thank you!
Just like I said above, if you have suggestions, maybe want to sponsor the fastest growing newsletter there currently is, please reply to the email.
I’m all ears!
Edward Frank Morris
The Context_
Please note that all images are AI Generated and are for illustration purposes only.
