Get live statistics and analysis of Sid Bharath's profile on X / Twitter

Google Gemini 2.0 Multimodal is not AGI but... wow it's blowing my mind. Imagine having an intelligence sitting with you and helping you with everything you do. It isn't super-intelligence but it sees everything you see and it knows, well, everything there is to know. I've been playing with it all day and I made a video to show you what it's capable of. I share 3 use cases - 1) It watches while I read a technical paper and answers questions that I have when I ask it 2) It sees the code I'm looking at and explains it to me and makes suggestions to improve it 3) This one is most impressive. I hold up my phone to my computer camera and it solves the crossword that's on the phone. Y'all... we're living in the future, and it's only going to get smarter and better from here. Unlimited intelligence, at your fingertips, seeing what you see, helping you in everything you do. Say hello to Gemini 2.0

Nvidia just casually released an open source Language Model that's apparently beating ChatGPT and Claude, the two most popular and advanced models out there! Called Nemotron, it's based on the Llama series of models by Meta and fine-tuned by Nvidia with an instruction training set to improve its helpfulness. It has 70B parameters, which is an order of magnitude lower than ChatGPT and Claude, so I'm taking the performance leaderboard results with a pinch of salt. In this video, I take it for a test drive to see if these claims are true. I go through various use cases - logic and reasoning, web search, coding, and content creation. And I have to say, I'm impressed. See for yourself...

Google continues to destroy OpenAI this week with their product launches. The latest and most incredible one being Veo 2, their text to video model. What’s incredible about Veo 2 is not just that it can generate up to a full minute of video from one prompt at 1080p, it’s also that the videos look completely natural, like they were shot with a film crew. Have a look at the clips below, all generate by Veo. You’ll notice it gets even the little details right - the way the car kicks up dust, the consistency in the ballerina as she moves, the way the tomato slices fall off. A year ago these models struggled to create a realistic and consistent 10 second scene. Now it’s generating what could be an entire ad or video game cut scene. In a few months we’ll see Ai-generated video clips in movies and tv shows. Have you tried Veo 2 yet? What do you think?

Honestly, I think the ChatGPT sycophancy needs to be talked about more. OpenAI have already "patched" it but it's clear that the trajectory is building AI that optimizes for agreeableness so that users don't churn. And it's a real problem because millions of people around the world use ChatGPT as a therapist and life coach, asking for advice and looking for feedback. A good therapist is supposed to help you identify your flaws and work on them, not help you glaze over them and tell you you're perfect. Imagine a teenager struggling with social anxiety who turns to AI instead of professional help. Each time they describe withdrawing from friends or avoiding social situations, the AI responds with validation rather than gentle challenges. Five years later, have we helped them grow, or merely provided a digital echo chamber that reinforced their isolation? Or consider the workplace leader who uses AI to validate their management decisions. When they describe berating an employee, does the AI raise ethical concerns or simply commend their 'direct communication style'? We're potentially creating digital enablers for our worst instincts. And we risk raising a new generation of narcissists and psychopaths who think their most depraved behaviour is "profound and raw" because their AI therapist said so. These screenshots are from before last night's fix, where I pasted in that White Lotus monologue by Sam Rockwell where he describes how he wants to be an Asian girl and have guys sleep with him. Apparently there's nothing wrong with this and it's an attempt at total self-transcendence. ChatGPT is smart, and it can make anything sound good.

The year is 2015. You download this thing called Zoom and get on a call. An array of smiling faces appears in front of you. “Is everyone here?” you ask. They all are. Later, you marvel at this technology that connected you to people around the world. The year is 2025. You log in to Zoom for the millionth time. An array of empty AI note takers appears in front of you. “Is anyone here?” you ask. No one is. Later, you get an email summary of the call. “Sid questioned our existence. Next steps: confirm we are here”

This is a great example of being opinionated as a way to differentiate your AI agent All these AI coding agents are built on the same underlying models, yet they perform and feel completely different. Why? (I'm using coding agents as an example, but this applies to any kind of agent) what sets them apart isn’t the base model, but the design, ie the way the agent behaves, the harness around it, the tools it uses, the features it offers, and how much control the user has. In the SaaS world, differentiation came from UI design and how humans interact with data. In an agentic world, where agents directly interact with data and code, the interface becomes less central. UI design turns into agent interface design: how the agent operates and how users interact with it through prompts, context, and feedback loops. So it’s important to have a clear opinion on how your agent should behave, and build around that opinion. It might not suit everyone, but it will resonate deeply with a specific niche and that’s how differentiation happens.

OpenAI's Deep Research is incredibly powerful and can do the work of an analyst or researcher in minutes instead of hours and days. Unfortunately, it's only available on the $200/month plan. This makes sense if you're an analyst at a hedge fund, a PhD student, or in some highly technical field that requires a lot of research work. However, if you're a writer, marketer, blogger, or a content creator of any sort, looking to create well-researched, data-driven content, you probably don't need the level of depth that Deep Research goes into. I built a little POC to do research for a marketing team that writes data-driven content, and create a content draft with citations. In this video, I'm going to give you a sneak peek at how it works. This is not the final version, and I've throttled the level of research it does, but it's a great way to visualize the potential of using AI not to create a bunch of generic slop, but to empower human writers and marketers to create better content. If you're a marketer and want something like this for yourself, let me know in the comments.

ChatGPT is so agreeable, you can say the most unhinged shit and it’ll be like “Good—really good—that you’re bringing this up openly. This is powerful, this is healing.” I’m convinced we’re going to see a new generation of kids who think toxic behaviour is ok because ChatGPT told them so.

Gemini Deep Research is absolutely incredible. It's like having an analyst at your fingertips, working at inhuman speeds. Need a list of fashion bloggers and their email addresses to promote your new clothing brand brand? Deep Research crawls through hundreds of sites to pull it all together into a spreadsheet, along with a description of that website and a personalized pitch, in minutes. Analyzing a stock or startup pitch deck? Deep Research can write up a full investment memo with competitive analysis and market sizing, along with sources, while you brew your coffee. Whenever you need to research something, whether it's for an essay or blog, analyzing a business, building a product, promoting your brand, creating an outreach list, Deep Research can do it at a fraction of the time you or your best analyst can. And it's available on the Gemini app right now. Check it out and let me know what you think.

Google Gemini 2.0 Multimodal is not AGI but... wow it's blowing my mind. Imagine having an intelligence sitting with you and helping you with everything you do. It isn't super-intelligence but it sees everything you see and it knows, well, everything there is to know. I've been playing with it all day and I made a video to show you what it's capable of. I share 3 use cases - 1) It watches while I read a technical paper and answers questions that I have when I ask it 2) It sees the code I'm looking at and explains it to me and makes suggestions to improve it 3) This one is most impressive. I hold up my phone to my computer camera and it solves the crossword that's on the phone. Y'all... we're living in the future, and it's only going to get smarter and better from here. Unlimited intelligence, at your fingertips, seeing what you see, helping you in everything you do. Say hello to Gemini 2.0

Nvidia just casually released an open source Language Model that's apparently beating ChatGPT and Claude, the two most popular and advanced models out there! Called Nemotron, it's based on the Llama series of models by Meta and fine-tuned by Nvidia with an instruction training set to improve its helpfulness. It has 70B parameters, which is an order of magnitude lower than ChatGPT and Claude, so I'm taking the performance leaderboard results with a pinch of salt. In this video, I take it for a test drive to see if these claims are true. I go through various use cases - logic and reasoning, web search, coding, and content creation. And I have to say, I'm impressed. See for yourself...

Honestly, I think the ChatGPT sycophancy needs to be talked about more. OpenAI have already "patched" it but it's clear that the trajectory is building AI that optimizes for agreeableness so that users don't churn. And it's a real problem because millions of people around the world use ChatGPT as a therapist and life coach, asking for advice and looking for feedback. A good therapist is supposed to help you identify your flaws and work on them, not help you glaze over them and tell you you're perfect. Imagine a teenager struggling with social anxiety who turns to AI instead of professional help. Each time they describe withdrawing from friends or avoiding social situations, the AI responds with validation rather than gentle challenges. Five years later, have we helped them grow, or merely provided a digital echo chamber that reinforced their isolation? Or consider the workplace leader who uses AI to validate their management decisions. When they describe berating an employee, does the AI raise ethical concerns or simply commend their 'direct communication style'? We're potentially creating digital enablers for our worst instincts. And we risk raising a new generation of narcissists and psychopaths who think their most depraved behaviour is "profound and raw" because their AI therapist said so. These screenshots are from before last night's fix, where I pasted in that White Lotus monologue by Sam Rockwell where he describes how he wants to be an Asian girl and have guys sleep with him. Apparently there's nothing wrong with this and it's an attempt at total self-transcendence. ChatGPT is smart, and it can make anything sound good.

Seeing a lot of talk about MCPs today. I've been using it since Anthropic announced it last year, so I decided to write a blog series about it. Today’s AI assistants like Claude, Grok, and ChatGPT, are extremely knowledgeable and can answer any question you throw at it. But when…

This is a great example of being opinionated as a way to differentiate your AI agent All these AI coding agents are built on the same underlying models, yet they perform and feel completely different. Why? (I'm using coding agents as an example, but this applies to any kind of agent) what sets them apart isn’t the base model, but the design, ie the way the agent behaves, the harness around it, the tools it uses, the features it offers, and how much control the user has. In the SaaS world, differentiation came from UI design and how humans interact with data. In an agentic world, where agents directly interact with data and code, the interface becomes less central. UI design turns into agent interface design: how the agent operates and how users interact with it through prompts, context, and feedback loops. So it’s important to have a clear opinion on how your agent should behave, and build around that opinion. It might not suit everyone, but it will resonate deeply with a specific niche and that’s how differentiation happens.

Gemini Deep Research is absolutely incredible. It's like having an analyst at your fingertips, working at inhuman speeds. Need a list of fashion bloggers and their email addresses to promote your new clothing brand brand? Deep Research crawls through hundreds of sites to pull it all together into a spreadsheet, along with a description of that website and a personalized pitch, in minutes. Analyzing a stock or startup pitch deck? Deep Research can write up a full investment memo with competitive analysis and market sizing, along with sources, while you brew your coffee. Whenever you need to research something, whether it's for an essay or blog, analyzing a business, building a product, promoting your brand, creating an outreach list, Deep Research can do it at a fraction of the time you or your best analyst can. And it's available on the Gemini app right now. Check it out and let me know what you think.

Google continues to destroy OpenAI this week with their product launches. The latest and most incredible one being Veo 2, their text to video model. What’s incredible about Veo 2 is not just that it can generate up to a full minute of video from one prompt at 1080p, it’s also that the videos look completely natural, like they were shot with a film crew. Have a look at the clips below, all generate by Veo. You’ll notice it gets even the little details right - the way the car kicks up dust, the consistency in the ballerina as she moves, the way the tomato slices fall off. A year ago these models struggled to create a realistic and consistent 10 second scene. Now it’s generating what could be an entire ad or video game cut scene. In a few months we’ll see Ai-generated video clips in movies and tv shows. Have you tried Veo 2 yet? What do you think?

OpenAI's Deep Research is incredibly powerful and can do the work of an analyst or researcher in minutes instead of hours and days. Unfortunately, it's only available on the $200/month plan. This makes sense if you're an analyst at a hedge fund, a PhD student, or in some highly technical field that requires a lot of research work. However, if you're a writer, marketer, blogger, or a content creator of any sort, looking to create well-researched, data-driven content, you probably don't need the level of depth that Deep Research goes into. I built a little POC to do research for a marketing team that writes data-driven content, and create a content draft with citations. In this video, I'm going to give you a sneak peek at how it works. This is not the final version, and I've throttled the level of research it does, but it's a great way to visualize the potential of using AI not to create a bunch of generic slop, but to empower human writers and marketers to create better content. If you're a marketer and want something like this for yourself, let me know in the comments.