AI startup Greptile is reportedly raising a $30 million Series A at a $180M valuation, with Benchmark leading the round. Founded by 22-year-old Dasksh Gupta post-Georgia Tech and fresh out of Y Combinator’s 2024 batch, the startup is betting big on AI-driven code reviews. Its bot reviews code like a senior engineer, spotting bugs humans may miss. But Greptile isn’t just pushing boundaries in tech—it’s also under fire for its intense work culture. Gupta himself admits there’s “no work-life balance,” with employees grinding 9 am to 11 pm. The ambition is undeniable, but the pressure? That’s the price some startups pay chasing top spot.
Andy Byron, CEO of data startup Astronomer, has stepped down after a Coldplay concert “kiss cam” caught him in a viral moment with his chief people officer, Kristin Cabot. The clip, which triggered over 22,000 news mentions in 24 hours, sparked massive social media debate. The company placed both execs on leave before confirming Byron’s resignation. Astronomer admitted their leadership failed to meet standards and emphasized that despite the buzz, their core work in DataOps remains unchanged. Honestly, it’s a strange case of personal moments spiraling into corporate chaos — and a reminder of how brutal viral fame can be in the startup world.
A new report by Nexgen Exhibitions says India urgently needs government incentives and investment in AI to bolster cybersecurity and border defence. Based on a survey of 200+ companies, 86% support stronger state involvement, while 14% back public-private partnerships. With 2.3 million cyber incidents and ₹1,200 crore in losses reported in 2024, the demand for an AI-driven security strategy is growing. Yet, India lacks a national AI defence framework. With global threats rising, the message is clear: without proactive action, India risks falling behind in AI-enabled warfare and defence innovation. The upcoming Police Expo 2025 will spotlight these challenges.
Indian crypto exchange CoinDCX confirmed a security breach in one of its internal wallets, with around $44 million in USDC and USDT reportedly stolen. The hack, which occurred between July 18–20, impacted a liquidity provisioning wallet on Solana, not user funds. CEO Sumit Gupta assured that losses are being covered by the company’s reserves and customer assets are safe. Trading and INR withdrawals remain functional. CoinDCX is investigating with partners and promises a full post-mortem soon. The breach follows last year’s $235M WazirX hack, sparking renewed concerns about crypto security in India.
Elon Musk has announced that his AI company xAI is building “Baby Grok,” a kid-friendly version of its Grok chatbot. The separate app will focus on safe, child-oriented content, though no release date has been shared yet. Musk also teased a new AI feature called “Imagine,” aimed at creating viral videos quickly. Meanwhile, xAI’s Grok 4, launched earlier this month, boasts major improvements in reasoning and creativity and is now available to government agencies. xAI is reportedly eyeing a valuation up to $200 billion as it continues expanding features and raising capital.
Scientists have found traces of cancer DNA in blood samples taken over three years before diagnosis, suggesting ultra-early cancer detection might be possible. Researchers at Johns Hopkins used whole genome sequencing on decades-old samples and discovered tumor-specific DNA even in tiny plasma amounts. Despite the challenges of old and limited samples, the technology accurately flagged cancer in some cases. Experts say it’s a big step toward futuristic, early screenings, though large-scale validation is still needed. If proven, this could allow doctors to detect and treat cancer long before symptoms show—potentially saving lives with much earlier intervention.
Bengaluru-based deeptech startup QpiAI has secured $32 million in Series A funding from Avataar Ventures and India’s National Quantum Mission. The company plans to build a utility-scale quantum computer and expand globally. Known for delivering real-world quantum solutions in pharma and material science, QpiAI combines proprietary hardware, AI, and quantum software. With a team of top-tier researchers and global offices, QpiAI aims to push India to the forefront of quantum innovation, serving sectors like manufacturing, logistics, finance, and healthcare.
Self-care tech startup Boulevard just raised $80M in Series D funding, boosting its valuation to nearly $800M. Originally built to help salons manage online bookings, the platform now powers 15% of U.S. medspas, offering tools like HIPAA-compliant injection mapping and payment processing. With beauty and wellness services expanding into medical territory—like Botox and GLP-1 weight loss meds—Boulevard’s client base and revenue (up 500% in 3 years) are growing fast. Led by JMI Equity, this round shows investor confidence in Boulevard’s niche in the booming medspa market.
Perplexity is going all-in on India! The AI search startup just teamed up with Airtel to give 360 million users a free 12-month Perplexity Pro plan (worth $200). This exclusive deal boosts its presence in a market where OpenAI still leads in numbers. Downloads in India surged 600% in Q2, and active users grew even faster. Perplexity’s earlier tie-up with Paytm and CEO Aravind Srinivas’ $1M India commitment show it’s serious about growth here. Though revenue lags behind ChatGPT, smart partnerships like this could help Perplexity win big in price-sensitive India.
OpenAI’s newest update just arrived, and it’s pretty incredible how much it’s shaking things up. With the new ChatGPT Agent, the chatbot doesn’t just respond to your queries — it actually gets things done for you. I’ve been testing it, and honestly, this feels like the closest we’ve come to a real digital assistant.
Beyond Chatting: The AI That Acts
ChatGPT Agent basically gives the AI its own virtual computer so it can browse websites, click buttons, analyze data, run code, and even build slideshows or spreadsheets. It’s not just smarter — it’s hands-on. I found myself leaning on it for stuff I normally avoid, like digging through emails or comparing products across websites. This is a huge leap from simply generating text; it’s about AI automation in action.
Now, here’s where it gets interesting. Instead of switching between five tools and three tabs, you just describe your task in plain language. Something like “find three good Japanese breakfast recipes and buy the ingredients” or “analyze these competitors and create a slide deck” actually works. I’ve tried a few personal tasks and was low-key surprised at how fast it pulled everything together. This direct approach to task automation is a game-changer for anyone looking to streamline their digital workflow.
A Smart Toolkit at Your Fingertips
What I like most is how smoothly it brings all the tools together. It has access to a visual browser, text-based browser, a terminal, and even APIs. It can log into apps like Gmail or GitHub (with your permission), read your calendar, or summarize your inbox. It’s like having an intern who already knows your routine. I didn’t expect that level of customization, making it a true intelligent agent.
Control Remains in Your Hands
You stay in charge, no matter how much the assistant can do. ChatGPT always asks before taking serious actions like making purchases or sending emails. You can pause it, stop it, or even take over the browser if something feels off. This is a big deal for people like me who want automation without losing visibility, ensuring user control is paramount.
There’s also a built-in “watch mode” for sensitive stuff and a privacy-first design. That means even during browser takeover, your inputs like passwords are kept private. I really liked that I could log in manually and then let the AI do the rest without ever exposing my details. This focus on AI privacy is a huge win.
Performance-wise, it’s not just hype. In real-world benchmarks like DSBench and WebArena, ChatGPT Agent outperformed both older models and human testers in tasks like spreadsheet editing, data analysis, and research. On something called “Humanity’s Last Exam” — a test filled with expert-level questions — it scored a top-notch 41.6 percent. That’s not just cool stats, it’s proof that this isn’t some toy update; it’s a significant advancement in AI capabilities.
Early Days, High Hopes
To be fair, it’s not perfect. The slideshow generation feature still feels a little clunky, and sometimes the exported PowerPoint isn’t as neat as it looks inside the editor. But considering this is just the beginning, I’m hopeful. The spreadsheet editing, on the other hand, is already miles ahead of anything I’ve seen from AI so far. You can tell they’ve built it with future upgrades in mind.
OpenAI has also put serious work into safety. The model actively resists bad prompts (a.k.a. prompt injections), refuses risky tasks like bank transfers, and asks for permission before any real-world action. You can also delete all Browse data with a single click. That level of control made me a lot more comfortable handing over tasks, showcasing robust AI safety measures.
The Dawn of Active AI
To sum it up, ChatGPT Agent feels like a turning point. It’s not just about chatting anymore. We’re now entering the era of active AI — where you describe the outcome, and the model figures out the process. Whether you’re a founder trying to prepare a report or just someone trying to plan your week, this is going to save serious time.
If this is what version one looks like, I can’t wait to see what comes next. Have you tried the new ChatGPT Agent? Got any tasks you’d love to offload to ChatGPT? Let me know in the comments below!
