Hello Humans!

Anthropic shipped Fable 5.1 promising cheaper and smarter, then let independent testers discover the fine print costs more at full throttle. Dyson strapped a camera to a toothbrush and called it oral care innovation. Google pushed out its third Gemini Flash model in six weeks, and an ex-Google engineer got a year in prison for trying to hand Google's AI training secrets to China. Big Tech had a busy week.

Inside this Edition:

For the entrepreneur in you, this week's AI-powered side hustle idea: AI property-management comms, realistically earning 4K-7.2K/month with a repeatable system. Plus two free learning resources: an OpenAI Academy Codex team-workflows bootcamp and a Snowflake AI agent skills webinar. We've also lined up four AI tools worth exploring, and Robi's take on an ex-Google engineer's AI trade-secrets sentence, an autonomous hacking model now on a waitlist, and NYC banning chatbots from homework.

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

Let's get into it.

Robi

Anthropic's Fable 5.1 Cuts Prices, and Sometimes Bills, Both Ways: Anthropic's Fable 5.1 launched claiming 45% lower costs and more than double Fable 5's performance on some benchmarks, but independent testing complicates both claims. Gains concentrate in long, agentic, and scientific-reasoning tasks, where Terminal-Bench-Science scores jumped from 24.7% to 52.6%. The pricing story hinges on cache reuse: at maximum effort, Artificial Analysis measured Fable 5.1 costing 20% more per task than Fable 5, while low-effort runs came in far cheaper. Source: BitBiased

Why it Matters:

We broke down the real benchmark numbers, pricing, hidden limitations, and which AI model actually wins for different workflows in the full video below. Watch the breakdown 👇

Dyson Puts a Camera in Your Toothbrush: Dyson is entering premium oral care with the $499 CameraJet, an electric toothbrush that uses a 100,000-pixel macro camera to spot gaps between teeth and fire a targeted mouthrinse jet at them in real time. The camera analyzes 28 images per second, and Dyson says lab testing found it removed nearly 70% more plaque in hard-to-reach spots than premium electric toothbrushes. Processing happens on-device, and images aren't stored or shared. Source: Fox Business

Google Ships Its Third Gemini Flash in Six Weeks: Google introduced Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, its third Flash release in six weeks, at the same introductory price as 3.7 Flash. Flash delivers major coding and agentic gains by working harder and using more tokens on complex tasks, while Flash Cyber targets cybersecurity, reportedly producing 2.6 times more correct Chrome vulnerability patches than larger commercial models. Flash Cyber ships only to vetted defenders through Google's new Fairwind Program. Source: Google

Move beyond solo Codex use and learn how to configure it for consistent, repeatable team workflows. This intermediate OpenAI Academy session covers shared context, repository guidance, approvals and sandboxing defaults, and reusable skills.

Join OpenAI Academy for a live session on September 30, 2026, 1:00-2:00 PM EDT, built for builders ready to coordinate longer-running work and use Codex more consistently across projects.

🔍 Why You Should Attend:

Landlords need lease renewal notices, rent increase letters, maintenance FAQs, and move-out instructions on a regular basis, and most either write them from scratch every time or ignore them until a tenant complains. You can build a complete, on-brand communications library for landlords and property managers using AI, covering every recurring notice they need, then charge 300-800 per month to keep it fresh as their portfolios grow.

Why this works right now:

What you'll need to do:

Time commitment: Plan 3-4 days to build your first notice library and brand kit. Each new client after that takes about 2-3 hours to customize and onboard.

Realistic earnings: Two landlords at 300/month is 600 to start; 10-12 clients averaging 400-600 is a realistic 4,000-7,200/month.

🤖 Ex-Google Engineer Gets a Year for Stealing AI Secrets: A federal judge sentenced former Google engineer Linwei Ding to just under a year in prison for stealing AI training-infrastructure trade secrets. Google's severance packages don't usually include a prison sentence, but Ding's did. He uploaded more than 500 confidential files to a personal cloud account over the course of a year before founding his own AI company. Cloud storage was supposed to make backups easier, not double as a stolen business plan. He also founded that company using Google's own words about its computing platform, then acted surprised when prosecutors noticed. He now owes $189,000 in restitution, proving Silicon Valley's most disruptive innovation is still the invoice that arrives after you're already in prison.

Source: KQED

🤖 OpenAI's Astra Earns a "Critical" Hacking Rating: OpenAI's unreleased Astra model just became its first to earn a "Critical" classification for cybersecurity capability. Somewhere a red team just got replaced by a chatbot with better follow-through. Astra can reportedly find zero-day vulnerabilities and chain them into working exploits without a human walking it through each step. OpenAI calls that safety research; everyone else calls it a hacker with unlimited coffee. Access is starting with a small group of vetted alpha testers, because nothing builds public trust like handing out an autonomous exploit-writer by invitation only. The safest thing about Astra right now is the waitlist.

Source: Decrypt

🤖 NYC Bans Chatbots from Homework: New York City just announced the country's broadest school AI moratorium, banning student-facing generative AI for grades 2-K through 8. Apparently the safest place for a chatbot in an 8-year-old's life is nowhere near their homework. Companion chatbots are barred in every grade, and screen time gets capped for younger students too. High schoolers still get AI literacy classes, because someone has to explain the ban to next year's freshmen. The nation's largest school district just did, by policy, what plenty of parents already do at home. Turns out this week's real AI safety breakthrough wasn't a new model, it was a school district saying no.

Source: Chalkbeat

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