AI News: Gemini UPGRADED, GPT-5.3 LEAKED, Claude Cowork, AI Doctors!

If you blink, you miss an AI release. Today, we got big moves from Google, Enthropic, and OpenAI. Starting with Gemini getting a whole lot more personal. So, let's get into it. Google

just introduced personal intelligence inside Gemini. A new feature that lets Gemini understand you, not just the internet. The idea is simple but powerful. You can now connect Gemini to your Google apps like Gmail, photos, YouTube, and search with a single tap. And Gemini uses that context to give more personal and more useful answers. This is launching as beta in the US for Google AI Pro and AI Ultra users, but it shows where AI assistants are clearly heading. Most assistants can retrieve information. Gemini's advantage is reasoning across your own data, combining emails, photos, and search history only when you ask it to. That means it can pull specific details from emails, understand information inside photos, combine text, images, and video context, and give a single tailored response. This isn't just memory. It's contextual intelligence. Google shared a great example to show why this matters. Someone needed new tires, but didn't know their tire size. Gemini didn't just look it up. It suggested different tire options based on driving habits, reference past family road trips found in photos, and pulled ratings and prices automatically. When asked for the license plate number, Gemini extracted it directly from a photo and even confirmed the vehicle trim using Gmail. That's the difference. Less searching and more doing. Now, the obvious question is privacy. Personal intelligence is off by default. You choose actually which apps to connect. You can turn it off anytime. And Gemini explains where its answers come from. Importantly, Google says Gemini does not train directly on your Gmail or photos. Your data is referenced to answer requests, not absorbed into the model. You can also regenerate responses without personalization or use temporary chats if you want zero context. This is Google moving Gemini from a chatbot to a true personal assistant, one that lives inside its own ecosystem. And this is just the beginning. Google says that this is coming into AI mode in search with broader rollout planned over time. Now the real question is who is going to do personal AI best Google open AI or anthropic? Let's move on. Next up

Enthropic just quietly rolled out something very important. It's called co-work and it brings cloud code style agentic AI directly into the cloud desktop app. No terminal, no setup, just task. Right now, co-work is available as a research preview for Maxplan subscribers on Mac OS with a wait list for everyone else. So, what exactly is co-work? Co-work is in chat. Instead of responding one prompt at a time, Claude can now take one multi-step longrunning task and execute them end to end. You describe the outcome, step away, and come back to a finished work product. You can do things like organize your folders, synthesize research, Excel spreadsheets with real formulas, PowerPoint decks, fully formatted documents. This is the same agentic architecture behind cloud code now packaged for everyday knowledge work. The biggest shift here is local execution. Co-work runs directly on your computer and can read from and write to your local files, coordinate multiple sub aents in parallel, work for extended periods without timeouts, deliver output straight to your file system. Claude plans the task, breaks it into subtasks, runs them inside a virtual machine, and shows you what is doing along the way. You can watch, steer, or just let it run by itself. Enthropic is positioning co-work for work that normally takes hours. Think about things like organizing hundreds of file in your downloads folder. Turning receipts into a clean expense report, synthesizing research from papers, notes, and web sources, creating Excel models with VLOOKUPs and formatting, building side decks from messy notes or transcript. This isn't AI helping you work. This is AI doing the work for you. Because co-work is aic and has internet access. Key points to keep in mind. It runs inside an isolated virtual machine. You control which files Claude can access. You review its plan before it runs, but it can make real changes to your files. Also worth noting is that right now there's no memory across sessions, no project support yet, and it's only on desktop Mac OS only. The app must stay open while the tasks run, and it uses more compute than normal chat, so usage limits matter. Zooming out, this is a clear signal. Enthropic is betting that the future of AI isn't just smarter chat, it's delegation. Gemini is becoming personal. Claude is becoming agentic. And we're starting to see AI move from answering questions to actually operating your computer, which is a big shift. Let's move to the next story. A known AI engineer just teased

that GPT 5. 3 internally came soon, citing what he says is a very reliable source. Nothing official from Open AI yet, so this is still speculation, but the details are interesting. According to the post, GPT 5. 3 is expected to ship with stronger pre-training and IMO gold level reasoning techniques. That matters because IMO style reasoning isn't about speed or polish. It's about handling hard multi-step problems without falling apart halfway through. If this is true, it suggests Open AI is pushing deeper on core reasoning ability and not just adding features or UI improvements. Stronger pre-training also points to a better base model, which usually means fewer logic errors, better generalization, and more reliability when prompts aren't perfect. So, while this is still just ATS, it fits a bigger pattern we're seeing. Google is going personal, anthropic is going agentic, and OpenAI may be going allin on reasoning depth. If GPT 5. 3 Garlic is closed, this might explain why things have been unusually quiet lately from OpenAI. So keep your eyes open for this one. For the last couple of days, it

feels like every major AI lab is making a move into healthcare. And the way each one is doing it tells you exactly how they think trust will be won. So the way Open AI is doing it is through something called chat GPT health. They're not positioning this as a diagnosis or treatment. Instead, it's about helping people understand symptoms, prepare better questions for doctors, make sense of lab results and medical language, and navigate health information without panic. The tone is important. Chat GPT Health is framed as supportive, educational, and assistive, not authoritative. Open AAI is clearly trying to enter health care from the patientf facing side first, where clarity, empathy, and accessibility matter more than raw clinical power. At the same time, behind the scenes, OpenAI is still focused on healthcare infrastructure, things like documentation, administrative workflows, research assistance, and clinical support tools. It's a slow careful approach, but I think that's very much on purpose. Enthropic, on the other hand, feels a little bit more like a surgical approach. Claude is already being used across healthcare and life sciences for things like summarizing clinical and regulatory documents, analyzing biomedical research, supporting drug discovery and trial design, and assisting with medical writing. What Enthropic keeps emphasizing is restraint. Claude is designed to avoid overconfidence, avoid hallucination certainty, and defer when things get ambiguous, which in healthcare is basically all the time. This isn't about replacing clinicians. It's about helping them think more clearly. Then there's Google and they're taking a very different path with Medgamma 1. 5. Google is saying some healthc care problems are too sensitive for general purpose models. Medgamma is purpose-built for medicine with a focus on medical image interpretation like radiology and pathology, structured clinical reasoning, and medical speech to text through med ASR. This is Google leaning into its strength data infrastructure and domain specific models rather than stretching Gemini to do everything. When you put this together, the picture becomes clear. Healthcare is becoming one of the most important providing grounds for AI. Open AAI is entering through patient experience and infrastructure. Enthropic is betting on trusted reasoning and safety. And Google is going deep with specialized medical models. Different strategies, but the same realization. If AI can earn trust in healthcare, it can earn it anywhere. And that's why this space is moving slower, quieter, and more carefully than consumer AI, not because progress is slow, but because the stakes are high. Make sure to

subscribe to our channel. We do real tests, not just headlines. Make sure you're also subscribed to the world of AI. And don't forget to check out our newsletter for deeper breakdowns you won't see on YouTube. And I'm growing my Twitter following, so make sure you follow me on Twitter as well. Hope you guys enjoyed today's video and I'll see you in the next