OpenAI Pulls Away from Microsoft’s Grip
Plus: Gemini CLI brings AI to terminals, ElevenLabs turns voice into action, Midjourney makes its AI video debut, and more.
Hello Engineering Leaders and AI Enthusiasts!
This newsletter brings you the latest AI updates in a crisp manner! Dive in for a quick recap of everything important that happened around AI in the past two weeks.
And a huge shoutout to our amazing readers. We appreciate you😊
In today’s edition:
⚔️ Microsoft and OpenAI clash over control
💻 Google drops Gemini CLI for developers
🎙️ ElevenLabs launches new voice assistant
🎞️ Midjourney rolls out its first AI video gen model
🧬 DeepMind builds AI for decoding DNA
📚 MiniMax drops 1M-token AI model
🧠 Knowledge Nugget: The AI Panic: What I'm Actually Worried About by
Let’s go!
Microsoft and OpenAI clash over control
Tensions are rising between Microsoft and OpenAI as the two companies renegotiate the terms of their six-year partnership. OpenAI wants more independence, seeking approval to go fully for-profit, limit Microsoft’s access to its tech, and expand beyond Azure. Microsoft, meanwhile, wants deeper access and a larger stake in OpenAI’s future.
Talks have become so strained that OpenAI has even considered accusing Microsoft of anti-competitive behavior. The disagreement also centers around OpenAI’s $3B acquisition of Windsurf, its Stargate data center, and long-term rights over its future models, including any that may cross into artificial general intelligence (AGI).
Why does this matter?
What was once AI’s most iconic alliance is now unraveling. If OpenAI escalates to legal action, it shows just how urgent its push for independence has become, especially with AGI ambitions looming and its biggest backer turning into its biggest rival.
Google drops Gemini CLI for developers
Google just launched Gemini CLI, a new command-line tool that brings its top-performing AI model Gemini 2.5 Pro directly into developers’ terminals. It helps with writing code, running commands, managing files, searching the web, and even generating images and videos. And it’s completely free for up to 1,000 daily queries.
The tool is open-source and designed to fit into everyday developer workflows, acting like a smart assistant that lives in your terminal. It even adapts to specific projects and supports advanced coding help through Google’s Code Assist integration.
Why does it matter?
The race to win over developers is heating up, and Google just played its free card. By offering generous usage limits and open-sourcing the tool, it’s making a clear push to pull devs away from paid tools like OpenAI’s and Anthropic’s while positioning Gemini as the go-to AI agent inside real coding workflows.
ElevenLabs launches new voice assistant
AI voice platform ElevenLabs has introduced 11ai, an experimental voice assistant that talks and acts. Unlike traditional assistants, 11ai connects to apps like Slack, Notion, Linear, and Perplexity, allowing users to complete real tasks with voice commands. It’s built on Anthropic’s Model Context Protocol, making it easier to integrate with other tools and even custom workflows.
With over 5,000 voice options and voice cloning capabilities, 11ai runs on ElevenLabs’ own infrastructure and is free to try for a limited time.
Why does it matter?
ElevenLabs already leads in speech models, but combining that with tool integrations through MCP shifts the game. 11ai shows how voice assistants can evolve from novelty to utility doing real work across apps, and not just answering questions like Siri.
Midjourney rolls out its first AI video generation model
Midjourney has launched V1, its first video-generation model that turns any image into a short, animated clip. The web-only tool supports both automatic animation and manual prompts for things like camera movement or object actions. Each job outputs four 5-second clips (extendable to 20 seconds) and costs 8x the price of a still image, which the company claims is still 25x cheaper than its competitors.
V1 supports images from both Midjourney and external sources, and the videos preserve the platform’s signature visual aesthetic. CEO David Holz sees this as a key milestone on the path to building real-time, open-world AI simulations in the future.
Why does it matter?
Unlike other video models chasing realism, V1 keeps Midjourney’s signature artistic style front and center. It’s not aiming to compete with tools like Veo 3 just yet, with no audio or text-to-video features, but it shows the company’s slow, deliberate move toward building its own version of the AI-powered holodeck.
DeepMind builds AI for decoding DNA
Google DeepMind introduces AlphaGenome, a new AI model capable of analyzing massive DNA sequences, up to 1 million base pairs to predict how genetic mutations influence thousands of molecular functions. Unlike older models focused on narrow tasks, AlphaGenome unifies multiple prediction tasks in a single tool, outperforming specialized models across genomic benchmarks.
In real-world tests, it helped researchers studying leukemia understand how certain mutations triggered cancer-causing genes. DeepMind trained it in just four hours using public datasets and half the compute of their prior model, suggesting this could become a powerful tool for scaling up biomedical research and diagnostics.
Why does it matter?
AlphaGenome gives researchers a scalable way to test how DNA mutations affect health, helping accelerate the hunt for disease-causing variants without waiting weeks for wet-lab results.
MiniMax drops a 1M-token AI model
Chinese AI startup MiniMax has open-sourced M1, a new reasoning model that boasts a 1 million-token context window, claiming it’s the world’s largest. The model is optimized for software engineering and agent-style reasoning, and it shows top-tier results in long-context benchmarks.
MiniMax trained M1 for just $535K using a custom algorithm called CISPO, which cut training time in half. That makes it one of the most cost-efficient large models ever released, especially given its performance across coding and complex reasoning tasks.
Why does it matter?
Chinese labs are making a serious impression in the open-source AI race. MiniMax’s M1 challenges the idea that top-tier models require massive funding, suggesting that smart design may now matter more than sheer scale.
Enjoying the latest AI updates?
Refer your pals to subscribe to our newsletter and get exclusive access to 400+ game-changing AI tools.
When you use the referral link above or the “Share” button on any post, you'll get the credit for any new subscribers. All you need to do is send the link via text or email or share it on social media with friends.
Knowledge Nugget: The AI Panic: What I'm Actually Worried About
In this article,
outlines three under-the-radar risks that concern him more than doomsday AI predictions: the decay of critical thinking, black-box AI systems no one can maintain, and the growing neglect of security. From CEOs telling teams to “just ask the AI” to engineers deploying code they don’t understand, he sees a cultural shift where speed trumps scrutiny.These habits are creeping into architecture design, legal decision-making, and even core business logic, without teams fully grasping how the systems behave or fail. At the root of it all? A loss of first-principles thinking. Allan argues that AI does not replace the engineering discipline; it demands more of it.
Why does it matter?
The real AI risk is shortcuts. If teams treat AI as magic, they’ll ship fragile systems and create blind spots they can’t fix. In a rush to be “AI-first,” many companies may be quietly building the next wave of unmaintainable tech debt.
What Else Is Happening❗
📱 Google launches Gemma 3n, a compact multimodal AI that runs on just 2GB RAM, powering real-time image, audio, and video processing on mobile devices.
🧠 MIT researchers introduced SEAL, a self-training AI framework where models generate their own data and improve autonomously, outperforming GPT-4.1 in key tasks.
👗 Google debuts Doppl, an AI try-on app that lets users visualize outfits on animated versions of themselves, powered by image uploads and AI-generated videos.
🤖 Anthropic’s Claude now builds AI apps from text prompts via “Artifacts,” letting anyone create and share interactive tools without touching code or API keys.
📝 OpenAI is reportedly developing ChatGPT productivity tools with Google Docs–style collaboration and multi-user chat, aiming to compete with Workspace and Office.
👁️ Reddit may adopt Sam Altman’s World ID Orb for anonymous but verified user checks, making it the first U.S. platform to test biometric proof of humanity.
🕶️ Meta teams up with Oakley to launch AI smart glasses for athletes featuring 3K video, longer battery life, and a high-profile campaign with sports stars.
📊 A Stanford study finds most workers want AI to assist, not replace, favoring automation of repetitive tasks over full takeover, especially in creative fields.
🧠 An MIT study found that students using ChatGPT for essays showed weaker brain activity and retention than those writing unaided or with Google search.
🛡️ OpenAI launched “OpenAI for Government” and signed a $200M DoD deal to develop AI tools for military and federal use including cyber defense and admin tasks.
⚡ Google graduates Gemini 2.5 Pro and Flash to production and launches Flash-Lite, a speedy, low-cost model with a 1M-token context window.
New to the newsletter?
The AI Edge keeps engineering leaders & AI enthusiasts like you on the cutting edge of AI. From machine learning to ChatGPT to generative AI and large language models, we break down the latest AI developments and how you can apply them in your work.
Thanks for reading, and see you next week! 😊