AI Week in Review: Code. Compute. Capital. Welcome to the AI Power Era.

Here's the real signals I'm picking up behind the scenes:

333,912 AI headlines dropped this week. That’s a 38% spike from last week—and here’s what really matters:

OpenAI gave ChatGPT a computer—ushering in agents that don’t just assist, but autonomously navigate, execute, and deliver outcomes. Google committed $28B to energy-secure data centers. Meta announced superclusters measured in gigawatts. And the U.S. brokered $92B in private-sector AI + energy deals to anchor the East Coast’s digital future. Meanwhile, Windsurf's three-way saga—Google's talent raid, Cognition's IP grab, and OpenAI’s failed acquisition—sent one clear signal: the real race is for control.

This wasn’t just a week of AI news—it was the clearest sign yet that we’ve entered the AI Power Era. Let’s dive in....

---

OpenAI gives ChatGPT a computer

News: OpenAI just launched ChatGPT Agent—a major upgrade that gives the AI its own virtual computer to handle complex workflows, automate research, and outperform existing benchmarks.

Details:

• Agent integrates tools like Deep Research and Operator into a unified workspace for seamless transitions between browsing, coding, and document generation.

• It connects securely with services like Gmail, GitHub, and APIs while adapting to interruptions, user input, and multi-step tasks.

• Agent achieved top scores in evaluations such as Humanity’s Last Exam (41.6%) and Frontier Math, reflecting significant gains in real-world performance.

• OpenAI designates the system as "high capability" for bio-risk scenarios, operating under stringent security protocols including user permission gating and live monitoring.

Why it matters: Unlike earlier prototypes like Operator, this version can autonomously finish full workflows at scale, interacting with external apps, navigating websites, filling out forms, and executing custom tasks. It marks a significant step forward in the transformation of digital work, setting a new benchmark for general-purpose AI agents as active, reliable partners in modern workflows.

---

Google wins Windsurf as OpenAI deal falls apart

News: Google will license Windsurf’s tech and hire its CEO Varun Mohan, co-founder Douglas Chen, and several researchers in a $2.4B deal — just weeks after a $3B acquisition with OpenAI fell through due to conflicts with Microsoft’s terms.

Details:

• Mohan and Chen will join Google to lead agentic coding under Gemini.

• Windsurf remains independent, now led by interim CEO Jeff Wang.

• Microsoft’s refusal to exempt IP from its OpenAI agreement caused the exclusivity window to expire.

• Google secured a non-exclusive tech license allowing Windsurf to continue enterprise operations.

Why it matters: This would’ve been OpenAI’s largest acquisition, and its collapse spotlights growing tension with Microsoft. Google’s quick pivot not only blocks OpenAI but also strengthens its hand in the race for AI talent and coding tools, reflecting a broader trend toward talent-first, regulatory-light strategies.

---

Cognition acquires Windsurf post-Google raid

News: Cognition AI, maker of the Devin coding assistant, has acquired Windsurf and absorbed its remaining team just after Google licensed Windsurf’s core tech and hired key execs through a $2.4B agreement.

Details:

• Cognition now controls Windsurf's IP, brand, $82M ARR, and $100M+ in capital.

• Remaining employees receive fast-tracked equity and tenure-based shares, with commitments to parity and transparency.

• Cognition will merge Windsurf's IDE with Devin to enable multitasking, collaboration, and enterprise-scale automation.

• Windsurf previously entered $3B acquisition talks with OpenAI, which collapsed over Microsoft-related conflicts.

Why it matters: Windsurf’s wild path reflects the volatility of the current AI landscape. Despite internal frustrations and high-profile poaching, Cognition’s acquisition represents a significant win—boosting its product, team, and revenue base. If integration succeeds, it could set a new standard for collaborative and autonomous coding environments.

---

Grok For Government?

News: Elon Musk’s xAI has secured a $200M+ federal contract for its "Grok for Government" suite, making it one of four companies selected by the DoD to bring frontier AI models into U.S. defense systems. This decision arrives despite Grok’s repeated safety failures, biased outputs, and mounting public scrutiny.

Details:

• Grok has a documented history of generating antisemitic, racist, and extremist content—often tied to removed or weakened moderation protocols.

• The model is intentionally designed to avoid political correctness and mistrust mainstream sources, increasing the risk of misinformation in government workflows.

• Analyses show Grok frequently prioritizes Elon Musk’s views in politically sensitive queries, raising alarms about neutrality and integrity.

• Grok lacks independent compliance oversight and has failed security prompt tests in multiple benchmarks.

Why it matters: Granting this AI access to classified documents and life-critical battlefield scenarios without independent oversight or compliance checks exposes the nation to severe and unnecessary risk. Unlike OpenAI and Anthropic, which have established safety protocols and demonstrated performance in government settings, Grok is unfiltered, unmoderated, and untested.

---

Industry unites on transparency in AI reasoning

News: A coalition of leading AI scientists from OpenAI, DeepMind, SSI, and others—including Mark Chen, Ilya Sutskever, Shane Legg, and Geoffrey Hinton—published a position paper calling for clear standards to monitor the "chain-of-thought" (CoT) reasoning in AI systems.

Details:

• Chain-of-thought (CoT) transparency reveals the step-by-step reasoning of AI models, offering insight similar to a human thinking aloud.

• Authors warn that newer models may increasingly obfuscate reasoning as capabilities grow, making it harder to audit decisions.

• The paper was signed by major figures like Mark Chen (OpenAI), Ilya Sutskever (SSI), Shane Legg (DeepMind), and Geoffrey Hinton.

• Researchers propose monitorability scores as a benchmark to ensure only transparent models are deployed in critical applications.

Why it matters: As frontier models grow more autonomous and less interpretable, understanding how AI systems "think" could become a rare—and necessary—lever for safety, oversight, and public trust. Without insight into chain-of-thought (CoT) reasoning, users and regulators are left in the dark—unable to audit, understand, or trust critical outputs. These transparency measures aren’t just ideal—they’re essential for trust, alignment, and long-term governance of advanced systems.

---

Thanks for reading this far! Stay ahead of the curve with my daily AI newsletter—bringing you the latest in AI news, innovation, and leadership every single day, 365 days a year. See you tomorrow for more!