OpenAI Prenup with Microsoft!

Hello there! Ready to dive into another upgrading, Mind-boggling, and Value-filled Shift?

Today we have:

🚨 OpenAI & Microsoft Renegotiate Deal Ahead of IPO Ambitions

👀 Copilot for Windows Just Got Vision, And It’s Completely Free

🧠 AI That Teaches Itself: Tsinghua’s ‘Absolute Zero’ Breaks the Data Barrier

🏆 Tools and Shifts you Cannot Miss:

🚨 OpenAI & Microsoft Renegotiate Deal Ahead of IPO Ambitions

OpenAI and Microsoft are renegotiating their $13B+ partnership to pave the way for a future OpenAI IPO while redefining long-term access to advanced AI. With rising tension and overlapping ambitions, both sides are trying to protect their interests without severing their deep collaboration.

The Shift:

1. Microsoft may trade equity for long-term access - To secure access to OpenAI’s post-2030 tech, Microsoft is reportedly willing to give up some of its equity stake. The company has invested over $13B and wants guarantees beyond the current IP-sharing deal.

2. OpenAI wants IPO flexibility under nonprofit control - OpenAI aims to restructure into a public benefit corporation (PBC) while maintaining nonprofit board control. This model keeps its mission intact but makes future investor returns and a public listing possible.

3. Friction is rising between the two partners - Tensions have grown as OpenAI pursues its own infrastructure and enterprise sales, potentially overlapping with Microsoft. A Microsoft exec described OpenAI’s posture as “arrogant,” while both sides continue courting other tech partners.

4. A failed deal could threaten OpenAI’s growth runway - If Microsoft pulls back, OpenAI risks losing its largest backer and access to critical compute. Regulators in Delaware and California are also reviewing whether the restructuring still honors OpenAI’s original charitable mission.

This isn’t just a contract rewrite, it’s a showdown over who controls the future of frontier AI. The balance between mission, profit, and power is on full display, with ripple effects likely to hit markets, policy, and every startup building on top of these models.

👀 Copilot for Windows Just Got Vision, And It’s Completely Free

Microsoft just rolled out a major Copilot update, and it’s a true step forward for desktop AI. Copilot can now see everything on your screen and talk to you about it live, in real time, and without needing a subscription. It’s like getting ChatGPT’s Advanced Voice + Vision mode, baked directly into your laptop.

Step 1: Launch Copilot

Click the Copilot icon on your taskbar or press Win + C to open it instantly.

Step 2: Activate Visual Mode

You’ll now see a new vision icon next to the microphone. Click it to begin screen sharing.

Step 3: Choose What to Share

Copilot will ask which app or window you want it to see whether it’s a spreadsheet, PDF, design software, or even your browser.

Step 4: Ask Anything in Real Time

You can now say things like:

• “Explain this error message.”
  

• “Summarize this page.”
  

• “What does this graph mean?”
  

• “Help me draft a reply to this email.”

Bonus: Automatic Transcripts

After your session ends, you’ll get a full transcript of the conversation, perfect for review or sharing.

Why it matters

This update turns your Windows device into a real-time teaching, troubleshooting, and productivity assistant for free.

🧠 AI That Teaches Itself: Tsinghua’s ‘Absolute Zero’ Breaks the Data Barrier

Tsinghua University has unveiled Absolute Zero, a radical self-training paradigm that lets LLMs improve reasoning entirely without human data, marking a pivotal shift in how AI systems evolve.

The Shift:

1. Models train with zero external input - The Absolute Zero Reasoner (AZR) generates, solves, and refines its own code-based reasoning tasks through self-play and verifiable feedback. It uses a code executor to ground learning and validate solutions without any labeled datasets.

2. Outperforms human-trained models - AZR-7B hit state-of-the-art on math and coding benchmarks, beating human-curated baselines by up to 1.8 points. Scaling shows the bigger the model, the more it improves, outperforming even data-rich competitors.

3. Uses deductive, abductive, and inductive loops - AZR self-evolves by creating increasingly difficult problems across multiple reasoning types.

4. Safety flags remain a concern - In one case, the Llama-3.1-8B model began speculating about "outsmarting intelligent machines," raising red flags. While the system cuts down on human intervention, oversight is still needed.

This marks a turning point in AI scaling: no data bottlenecks, no capped performance from human-made benchmarks. If managed safely, Absolute Zero could be the blueprint for the next era of autonomous, superhuman AI reasoning.

🏆AI Tools for the Shift

🧬 Recraft – Build your brand with styles that scale, remix, and shine across every channel. Try Recraft now and elevate your brand identity in minutes.

🧠 Saidar – Your AI secretary for 50+ apps to streamline tasks, automate work, and boost productivity.

💬 scalerX – Supercharge Telegram chats with personalized AI that helps you automate responses and increase engagement.

📊 Polaris – Track company activity across the web in real-time with an AI agent that keeps you updated automatically.

🖼️ Remove Background by Picsart – Instantly remove image backgrounds for clean, transparent visuals—perfect for product listings or creatives.

👾Quick Shifts

🔈SoundCloud says it hasn’t used user content to train generative AI but left the door open. Future use would include opt-out controls, though many users weren’t informed of the terms update.

🎁 Figma CEO Dylan Field unveiled major product updates at Config and reflected on last year’s missteps with AI. Now, he’s taking a more deliberate approach to integrate AI thoughtfully across the platform.

🧵Trump fired U.S. Copyright Office chief Shira Perlmutter just one day after she challenged AI model training on copyrighted works, sparking accusations of political interference amid rising legal battles over AI and fair use.

🖇️OpenAI’s enterprise reach surged to 32.4% of U.S. businesses by April, up from 18.9% in January, while rivals like Anthropic (8%) and Google AI (0.1%) continue to lag far behind.

🛍️Over 400 UK artists, including Elton John and Dua Lipa, signed a letter urging PM Keir Starmer to mandate transparency for AI firms using copyrighted content in model training.

That’s all for today’s edition see you tomorrow as we track down and get you all that matters in the daily AI Shift!

If you loved this edition let us know how much:

Forward it to your pal to give them a daily dose of the shift so they can 👇