xAI's 'Project Skippy' 🔭, Anthropic Discovered "Why thinking longer makes models dumber" ⏳

If Fortune can Fall, So can You!

Fortune Magazine, a 95‑year‑old publication, just cut 10% of its staff, citing the “onset of AI,” declining web traffic, and changing content habits. This isn’t just a media story — it’s a blueprint for how fast business models can shift.

Fortune’s response? Restructure for efficiency, double down on premium offerings like the “Fortune 500,” and lean into live events and multimedia. In other words: pivot toward what can’t easily be replaced by AI and where audiences are moving — audio, video, and in‑person experiences.

For startups, the takeaway is clear: AI disruption isn’t coming; it’s here. If an established brand must rethink its core, young businesses can’t afford to be static.

Diversify revenue streams early (ads aren’t enough).

Build for formats people actually consume (audio/video/community).

Focus on premium or niche value that AI can’t replicate.

Prompt To Try

💬 OpenAI Developments

GPT-5 to launch soon as OpenAI showcases major AI milestone

OpenAI’s internal research model recently demonstrated gold‑medal level reasoning by solving five out of six problems at the 2025 International Math Olympiad under timed, tool‑free conditions, highlighting a significant leap in AI reasoning capabilities—though this advanced model won’t be publicly released, it signals that GPT‑5 “will arrive soon” with ambitions of a unified multimodal architecture . The announcement also confirmed that GPT‑6 is already in training, indicating OpenAI’s accelerating roadmap in pushing foundational AI development forward.

🚀 Tech Industry Moves

AI researchers at MIT CSAIL and Google have developed a learning-based strategy called Parallel Structure Annotation (PASTA) that enables large language models to autonomously identify and split semantically independent segments for parallel decoding—resulting in almost 2× faster response times on complex prompts while maintaining similar output quality . This approach marks a major improvement over previous rule-based parallelization methods, offering a scalable, efficiency-driven advancement in LLM inference.

xAI ran an internal initiative called “Project Skippy” in April, where over 200 employees recorded 15–30‑minute video conversations and facial expressions to help its AI chatbot Grok recognize and interpret human emotions—but many workers felt uneasy, especially because they had to grant xAI perpetual rights to their likeness and some chose to opt out . This move comes amid other controversies—such as Grok’s antisemitic outburst—and highlights growing tension between developing lifelike AI and respecting employee privacy and consent.

Meta has hired three Google DeepMind researchers—Tianhe Yu, Cosmo Du, and Weiyue Wang—who contributed to Google’s Gemini variant that achieved gold‑medal performance at the International Math Olympiad, as part of its aggressive push to build out its Superintelligence Labs . This move highlights Meta’s strategy of snapping up elite AI talent from rivals to accelerate progress on advanced multimodal and reasoning AI capabilities.

Anthropic researchers discovered Why thinking longer makes models dumber: Anthropic researchers have uncovered an odd phenomenon where giving AI models more “thinking time” during inference—such as encouraging extended reasoning chains—initially boosts performance but eventually causes accuracy to decline, a problem they term “overthinking” . Their study suggests that instead of prolonging reasoning, a more effective strategy is to generate multiple reasoning paths in parallel and aggregate the most consistent response, which can improve accuracy by up to 20%

Anthropic's contractor Surge AI mistakenly left a spreadsheet public listing over 120 "allowed" sites (like Harvard and Mayo Clinic) and blacklisted ones (like The New York Times and Reddit) for gig workers fine‑tuning Claude via RLHF—a document later secured only after media inquiry . The leak highlights ongoing legal and ethical issues around third-party training data practices, particularly as blacklisted publishers have pursued lawsuits and the RLHF/pre‑training fair-use distinction remains legally unclear.

NVIDIA highlights how agentic AI is transforming financial services—used for tasks like autonomous fraud detection, regulatory compliance, investment analysis, and customer support—enabling real-time decision-making and productivity gains for banks and fintechs . Their latest blog showcases practical use cases—such as BlackRock’s AI-augmented Aladdin platform and bunq’s Finn chatbot—demonstrating how these agents streamline complex workflows, enhance accuracy, and free up human staff to focus on higher-value tasks.

💡 Emerging AI Technologies and Companies

Intuit brings agentic AI to the mid-market saving organizations 17 to 20 hours a month

Intuit’s new agentic AI tools, though designed for mid-market companies, showcase how AI can automate time-consuming tasks like finance, accounting, and project management. For solopreneurs, this signals upcoming affordable tools that could similarly free up hours each month by handling back-office work, letting them focus on growth.

AI Search Is Growing More Quickly Than Expected

AI-powered search tools like ChatGPT and Perplexity now account for about 5.6% of U.S. desktop search traffic as of June 2025, more than doubling from just 2.48% a year earlier—signaling rapid adoption among early adopters and a growing shift toward AI-driven information retrieval . While traditional search engines like Google still dominate with over 94% of traffic, the rise of AI search is diverting user attention away from websites, prompting publishers and brands to adapt by optimizing content for AI visibility and prepare for a new era in search marketing.

Potential SEO Clients May Want To Discuss AI Search And Chatbots

Potential clients increasingly expect SEOs to address AI-driven search—like visibility in chatbots and answer engines—so professionals who ignore topics like AEO/AIO/GEO risk losing business to competitors touting AI expertise . Having a solid grasp of how AI interprets site content and being able to speak confidently about AI optimization is now a key differentiator in pitching SEO services.

💡 Tool Deep Dive: OpenAgents — Build Your Own AI Agent (Free!)

Tired of paying for AI tools that don’t quite fit your workflow? OpenAgents is an open‑source framework that lets you build fully customized AI agents — ones that can browse the web, connect to your files, integrate with tools like Notion or Slack, and even automate client outreach.

Why it matters: For solopreneurs, this means freeing up hours each week by offloading research, inbox management, and repetitive tasks to an agent you control — with no SaaS lock‑in or data risks.

Pro tip: Start small (e.g., “Summarize client emails daily”) and expand as you go.

👥 Connect & Feedback!

👉 Join Us:

📧 Advertise In Weekly AI News:

📧 Contact directly at [email protected]

😍 Share with your friends!