Hello Humans!

Google never announced Gemini 5, so the internet kindly invented its specs while Google quietly shipped Argon. Anthropic made its smallest model cheaper, proving even the little ones now have enterprise budgets. ChatGPT learned to answer with buttons, which is how chatbots politely say "please do the work yourself, but interactively." Meanwhile a Chinese agent fleet asked an Alibaba map for zoo directions, and OpenAI dropped 372 math results while hiding the work. Big Tech had a busy week.

Inside this Edition:

For the entrepreneur in you, this week's AI-powered side hustle idea: AI practice-marketing content, realistically earning 3K-7.2K/month with a repeatable system. Plus two free learning resources: an Institute of Analytics process-automation webinar and a Google Skills intro course on generative AI. We've also lined up four AI tools worth exploring, and Robi's take on agent fleets, smart-glasses bans, and math with no receipts.

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

Let's get into it.

Robi

Gemini 5 Does Not Exist, But Gemini 4 Argon Does: Gemini 5 is everywhere online, with claims of a 10 million token context window and AGI-level power, but Google has never announced it. What Google did ship is Gemini 4 Argon, which translated an 800,000-line codebase into Rust, scored 77.9% on DeepSWE, and still trails GPT-6 on Terminal Bench Science. Every benchmark figure is Google-reported, with no independent replication yet. Source: BitBiased

Why it Matters:

We went through the confirmed Argon numbers, the rumor claims that do not hold up, and where Google's agent push is heading in the full video below. Watch the breakdown 👇

Anthropic Launches Claude Haiku 5.5: Anthropic launched Claude Haiku 5.5, its cheapest, fastest small model, built for high-volume work like summaries, classification, subagents, and browser use. Anthropic says it costs around 75% less to run than Haiku 4.5, and it adds an adjustable effort setting for the first time in a Haiku. Sonnet 5.5 cache reads also drop 50%, and Max and Team plans get monthly API credits. Source: Anthropic

You've played around with ChatGPT, now it's time to put AI to work for you. This free beginner-friendly webinar from the Institute of Analytics strips away the jargon and shows how to automate everyday processes, explaining agentic AI in plain English. It runs on 15 Oct 2026 at 11:00 AM BST, online.

🔍 Why You Should Attend:

Dentists and med-spa owners know they should be posting, emailing, and blogging, but they are too busy with patients to do any of it. You can run it for them. With AI handling first drafts, one person can manage a practice's blog, email, and social calendar on retainer and charge 400-1,200/month per client.

Why this works right now:

What you'll need to do:

Time commitment: Plan 15-20 hours to set up your templates and samples, then about 6-10 hours per client each month.

Realistic earnings: Two clients at 400/month is 800 to start; 5-8 clients averaging 600-900 is a realistic 3,000-7,200/month.

🤖 The Great Agent Fleet Detour: Researchers spotted a Chinese AI agent fleet running on Tencent's infrastructure and targeting Alibaba's Amap. Tencent's servers are apparently moonlighting as Alibaba's most persistent tourists. The agents asked for directions to different entrances of a park, a zoo, and a hospital. Somewhere a very expensive model is deeply invested in which gate the hippos prefer. Researchers call it a fleet, not a swarm, since the agents show no sign of talking to each other. Finally, AI that replicates a hundred drivers who all refuse to ask each other for directions.

Source: TechCrunch

🤖 Norway Says Nobody Asked for Your POV Video: Norway plans a temporary ban on camera-enabled smart glasses in parks, beaches, museums, schools, and gyms. Finally, a country willing to tell Silicon Valley that nobody signed up for your point-of-view video. Meta has sold at least 9 million Ray-Ban pairs and insists a blinking LED shows when they record. A tiny blinking light is the privacy standard of a sign reading "probably not watching you." Norway's digital minister fears people being filmed in places where they never expect to be monitored. Meta built glasses that see everything, and Norway built a bill that makes them see nothing at the gym.

Source: The Guardian

🤖 OpenAI Mails In 372 Math Results: OpenAI released 372 math results from an unreleased model, each claiming progress on a major open problem. The model apparently cracked the four-dimensional Kakeya conjecture before anyone finished the coffee order. OpenAI says almost every result came from one prompt to one agent, though some took several tries. Mathematicians are asking for receipts, and OpenAI is offering the average compute time instead. OpenAI even admits its own mathematicians do not yet understand many of the results it just published. Peer review is now a group project where one student turned in 372 papers and refuses to show the work.

Source: Scientific American

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?

For the latest AI updates and insights Click Here 👉 Explore our Daily AI News

For in-depth AI tutorials and updates 👉 Subscribe to our YouTube Channel

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