Hello Humans!

OpenAI released GPT-5.3 Instant promising fewer hallucinations, which means it will now invent academic sources with slightly more confidence. Google answered with Gemini 3.1 Flash-Lite, proving the real AI breakthrough is making sure the demo doesn’t bankrupt the startup. And Meta is testing an AI shopping assistant inside your chats, because apparently ads weren’t invasive enough until they could join the conversation. Big Tech clearly spent the week turning everyday apps into slightly smarter, slightly weirder assistants.

Inside this Edition:

For the entrepreneur in you, this week's AI-powered side hustle idea: launching an AI-powered market research service that helps startups track trends, competitors, and consumer behavior using tools like Perplexity and Glimpse, realistically charging 3K–8K per report and scaling to 15K–40K/month with subscription clients. Plus two free learning resources: a practical HubSpot guide to getting more done with ChatGPT, and a live session from OpenAI showing how AI can help refine resumes and practice interview questions with confidence. We’ve also pulled together four AI tools professionals are quietly using to save hours every week, along with a few AI stories proving the machines are getting bolder by the day.

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

Let's get into it.

Robi

OpenAI Launches GPT-5.3 Instant : OpenAI has released GPT-5.3 Instant, an update to the most widely used model powering ChatGPT. The new version delivers more accurate answers, smoother conversational flow, and clearer responses for everyday tasks such as writing, research, and problem-solving. It also reduces unnecessary refusals and defensive language, while lowering hallucination rates and improving reliability. GPT-5.3 Instant is now available in ChatGPT and via the API. Source: OpenAI

Why it Matters:

Google Introduces Gemini 3.1 Flash-Lite for Faster: Google has launched Gemini 3.1 Flash-Lite, a lightweight AI model designed to deliver high-speed responses at significantly lower cost for developers and applications that require efficiency at scale. The model focuses on fast inference and reduced resource usage while maintaining core Gemini capabilities. It expands Google’s Gemini 3 family and is now available through developer APIs and cloud platforms for building responsive AI-powered products. Source: Google

Meta Tests AI Shopping Assistant Across Its Apps: Meta is testing an AI-powered shopping assistant across Facebook, Instagram, and WhatsApp to help users discover products through natural language conversations. The tool suggests items from brands and advertisers, allows comparisons, and refines results with follow-up prompts. Powered by Meta’s AI models and advertising systems, the feature aims to turn social feeds and chats into interactive shopping experiences while expanding the company’s digital commerce ambitions. Source: BloomBerg

Explore practical ways to use ChatGPT to prepare for career opportunities, from refining your resume to practicing interview questions and improving confidence during the job search process.

This short live session demonstrates how AI can help you organize your experience, turn rough notes into clear resume bullets, and simulate interview practice so you can approach applications with greater clarity and readiness. The webinar is designed to be approachable whether you are actively applying for jobs or simply preparing for future career opportunities.

🔍 Why You Should Attend:

Traditional market research firms charge 50,000+ for insights that you can now deliver in days using AI tools that analyze patents, social media trends, and consumer behavior at scale. Startups and mid-sized companies are desperate for this intelligence but can't justify six-figure consulting fees, creating a wide-open opportunity for solo operators who know how to harness the right technology.

Why this works right now:

What you'll need to do:

Time commitment: 15–20 hours to build your research infrastructure and templates; 6–10 hours per custom research report once operational.

Realistic earnings: 3,000– 8,000 per custom research project. With 3–5 subscription clients paying monthly, you're looking at 15,000– 40,000/month.

Want the complete step-by-step guide with tool setups, pricing strategies, and launch tactics? Dive into the full breakdown below.

🧮 AI Cracks Math Puzzle: Cursor's AI agent ran unsupervised for four days and cracked a math research problem better than the official human answer. Unsupervised, unprompted, and apparently overachieving. The challenge was designed by top mathematicians specifically to test real research ability ,the kind humans spend careers on. Experts are cautiously calling the result correct, like a professor who can't bring themselves to give the robot an A. The robot got the A anyway. At this point, "peer review pending" just means the humans need a minute to update their resumes.

Source: X Post

⚖️ Musk Pushes AGI Claim: Court documents from the Musk vs OpenAI trial reveal that OpenAI leaders internally considered GPT-4o to be AGI and then quietly retired it without telling anyone. So AGI arrived, clocked in, and got laid off before the press release. Musk's filing argues GPT-4o falls outside Microsoft's license and that a miniaturized version was quietly funneled into Altman's own private biotech investment. OpenAI may have achieved the most significant milestone in human history and responded by hoping nobody checked the filing cabinet. They achieved AGI, hid it, retired it, and cut their CEO a slice, all before lunch. Turns out "safe and beneficial AGI" was never a mission statement, it was an alibi.

Source: X Post

🤖 Block Cuts Humans: Jack Dorsey announced Block is cutting 40% of its workforce, explicitly blaming AI for making those roles redundant. Most CEOs would at least pretend it's about "restructuring." Dorsey told investors "intelligence tools" have fundamentally changed how companies are built, and predicted most businesses would reach the same conclusion within a year. He also admitted Block had ballooned to 13,000 employees during the pandemic, a mistake he's now correcting with extreme prejudice. So this wasn't really AI taking jobs, it was AI taking the blame. AI didn't disrupt the workforce. It just gave Dorsey a cleaner story.

Source: CNBC

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