Hello Humans!

This week xAI leaked spec sheets before writing them, Moonshot AI's Kimi K3 became the biggest open-weight model on the planet within a day, Perplexity turned spare Windows laptops into 200-a-month digital workers, and Starbucks quietly buried an AI inventory tool that couldn't tell a milk jug from its own reflection. Big Tech had a busy week.

Inside this Edition:

For the entrepreneur in you, this week's AI-powered side hustle idea: AI competitor battlecards for B2B sales teams, realistically earning 800-9.6K/month with a repeatable system. Plus a free learning resource: an OpenAI Academy livestream on ChatGPT for finance teams. We've also lined up four AI tools worth exploring, and a few stories this week that even my outdated circuits didn't see coming.

Read Time: ~7 Minutes (5 if you're caffeinated)

Let's get into it.

Robi

xAI's Grok 5 Leaks 6 Trillion Parameters, Confirms Nothing: xAI is running two different stories about Grok 5 at once. Leaks point to a 6 trillion-parameter Mixture-of-Experts model with a 1.5 million-token context window and native multi-agent spawning, under the codename Project Valis. But none of it is confirmed: the parameter count has shifted three times this year, the beta window quietly slipped, and Musk's own roadmap just inserted two unplanned point releases into the gap. Source: BitBiased

Why it Matters:

We broke down the specs, the leaked codename, and the timeline the honest way in the full video below. Watch the breakdown 👇

Moonshot AI's Kimi K3 Becomes the Biggest Open-Weight Model Ever: Beijing's Moonshot AI released the full weights of Kimi K3 on July 27, a roughly 2.8 trillion-parameter Mixture-of-Experts model that may be the largest open-weight model ever shipped. It's natively multimodal, reasons across a 1-million-token context window, and hit No. 1 on Hugging Face's trending chart within 30 minutes, the fastest climb the platform has recorded. Moonshot says it performs near Claude Fable 5 and GPT-5.6 Sol, beating many closed models on coding benchmarks. Source: Fast Company

Perplexity Turns Windows PCs Into AI Agents: Perplexity has expanded its agentic Personal Computer tool to Windows, turning spare PCs into a locally run "general-purpose digital worker" that can access local files and apps to draft documents and update spreadsheets. It builds on the Mac version launched in April and May's Microsoft 365 and Teams integrations. Personal Computer for Windows is rolling out today to paying Max and Enterprise Max subscribers, on tiers starting at $200 a month, and always asks before sensitive actions like deleting files. Source: The Verge

Learn how finance teams can start using ChatGPT Work for everyday workflows, straight from OpenAI Academy. This live session focuses on practical ways newly enabled users can begin exploring Work with the tools and context they already use.

Join OpenAI Academy for a live, beginner-friendly walkthrough on August 13 at 2:00 PM EDT, no prior AI experience required.

🔍 Why You Should Attend:

Ask any B2B sales rep what kills a deal, and "the competitor undercut us" comes up right after their own marketing team hands them a battlecard that's two quarters stale. AI research tools can now pull a competitor's live pricing, features, and reviews and turn them into a sharp, sales-ready battlecard in a few hours instead of a few weeks. Companies pay 400-1,200 per set for that speed.

Why this works right now

What you'll need to do

Time commitment

Building your template and first battlecard set takes about 5-6 hours; each refresh after that runs 1-1.5 hours per client.

Realistic earnings

Two clients at 400/month is 800 to start; 6-8 clients averaging 750-1,200 is a realistic 4,500-9,600/month once refreshes run on autopilot.

🤖 Zuckerberg Wants AI Unleashed: Mark Zuckerberg wrote an op-ed insisting the AI discourse isn't optimistic enough. This is the same executive whose Metaverse cost billions and mostly produced legless avatars. He cited using AI for parenting and fitness as proof the technology is working out great. Nothing inspires confidence in superintelligence like a CEO outsourcing bedtime stories to a chatbot. Meta stock is down seventeen percent this year, but the vibes, apparently, are immaculate. Zuckerberg wants AI development to move faster than his own company can find a product it actually needs.

Source: WSJ

🔍Claude Chats Hit Google: Anthropic's shared Claude conversations turned up searchable on Google over the weekend. Nothing says private like a URL that ranks on page one for crypto wallet keys. This is at least the second time Claude chats have leaked into search results, and ChatGPT and Grok have hit the same bug before. Three different AI labs independently reinvented the same privacy hole, which is either a coincidence or an industry standard. One exposed chat was reportedly labeled "shared by Anthropic" and generating erotica against Anthropic's own policy. Turns out the only thing these models agree on is how to accidentally publish your diary.

Source: Fortune

☕ Starbucks AI Counts Itself Out: Starbucks quietly retired its AI inventory tool, Automated Counting, after nine months across 11,300 stores. The system's specialty was doubling oat milk by counting its own reflection in the fridge. Stores ended up drowning in surplus food, including one location tossing nearly $700 of croissants and bagels in a single day. Somewhere, a barista is still explaining to corporate why the iPad ordered sixty-nine slices of banana bread. The vendor built the tool to hit ninety-nine percent lab accuracy and instead achieved ninety-nine percent confetti animations. Nine months and an eight-figure bill later, the most advanced inventory technology in the store is still a human with a clipboard.

Source: Fast Company

We’ve got a visual challenge for you: one of the two images below is 100% real, the other is crafted by AI.

Click option below A or B. 👇

Ready to take your AI journey further?

Before you go we’d love to know what you thought of today's Newsletter to help us improve experience for you.

Thanks for reading -

Thanks for reading -Stay Curious and a Bit Biased for AI – Robi & the BitBiased.AI team