AI Coding Finally Fixed, New AI Models, Huge Gemini Deal and More!

You've been working with different agents from Claude, OpenAI, and Google. And of course, each of these agents has their own perks. Well, at the annual GitHub Universe Conference 2025, the company just announced Agent HQ. Instead of juggling with multiple agents, why not have a single command center where you can manage all of them? Hey everyone, if you're new here, this is AI Labs and welcome to the very first episode of Context Weekly where a single episode is going to inform you about everything that has happened in this past week with AI. Coming back to agent HQ, it lets you manage your agents by acting as an orchestrator between all your favorite ones, allowing you to use and control agent workflows easily within GitHub. It also provides enterprisegrade admin controls that give you authority and full control over the workflow. The coolest part of the agent HQ announcement was the idea that much of today's AI generated code is sloppy and not highquality enough for production. GitHub aims to solve this by introducing code review agents and testing features to help produce senior developer level code that's properly tested and ready for large-scale production. It's a platform where policies and code quality scoring are used to ensure that only highquality code makes it to production. It also features an integrated plan mode, a tool that helps you plan by asking clarifying questions and then defining the necessary steps. It then uses that plan to send properly formatted instructions to the agents, ensuring they know exactly what to do. It also provides dashboard metrics for senior developer level management. The coolest part is the broader vision behind agent HQ. Why limit repositories to just your code? Let's have repositories that include documents, files, and everything else you can imagine. These multi-dimensional repositories can greatly enhance the way your organization works with each department like marketing, sales, or finance. Having its own repository and agents managing their operations seamlessly. Agent HQ features a mission control interface that allows you to assign, track and control all your agents, not only through the web version, but also through your IDE, mobile, or CLI. You can also manage branches, handle merge conflicts, and oversee agents and their tasks with just one click, giving you the freedom to fully monitor and control your agent operations. Announced for release in the coming months, Agent HQ marks a major step forward, moving from the core code completion phase of AI to fully agentic and multimodal workflows. Before we move on to the next update, here's a word from our sponsor, Appify. Appify is your go-to platform for web scraping, data extraction, and large-scale workflow automation. Whether you're gathering market insights, training AI models, or powering growth research, Apify lets you collect structured data from any website or API without worrying about servers or scaling. Its serverless actor system helps you build and deploy workflows in minutes. And the Appify Marketplace acts like an app store where developers, automation builders, and vibe coders can share and monetize their tools so anyone can use them. And here's something big. Appify just launched the $1 million Appify challenge, inviting creators to build and publish new actors on the marketplace. Earn from every user, showcase your tools to a global audience, and compete for $1 million in total rewards. Register for the challenge on the Appify landing page and join our Discord community to attend an exclusive QA session with the Appify team this Friday. Click the link in the pinned comment below to get started. The next huge update came from Cursor, which introduced Cursor 2.0 along with its new Composer model. The Composer model is an incredibly fast and capable system. They claim it outperforms existing models, showing impressive results in terms of both intelligence and speed. They introduced a new agent layout which features a chat interface that lets you interact with all their models including the new composer one which comes with a 200k context window. Not only that, you can also use multiple models at the same time, choosing whichever ones you want and comparing their answers and evaluating their performance side by side. The interesting feature that came with this update is the native browser support which allows you to see your code in action side by side with the coding environment. The cool part is that if a feature doesn't work the way you want, you can simply select it and ask cursor to make further changes. It also provides Chrome Dev Tools integration, allowing you to use them directly inside Cursor. This update gave Cursor room to make a strong comeback in the competition among AI coding assistants. In the competition among AI models, Google gave us another update. The company plans to release the Gemini 3 Pro preview in November. We all know that Gemini's previous models 2.5 Pro and 2.5 Flash have already performed really well in terms of both capability and their large context windows. The new model is expected to feature a 1 million token context window and will be specialized for general use cases. This also ties in with recent news that Nano Banana 2 will be coming out soon, an improved version of Nano Banana, Google's top performing image generation model. So, this is definitely something exciting to look forward to in November. Just like Google, OpenAI has also decided to step up the game by hinting at the release of GPT 5.1. We all know that GPT5 hasn't been performing too well according to recent reviews and user feedback. However, some new mystery models have been uploaded to Design Arena, a web-based platform where models compete in design with each other and their performance is evaluated by users. These new mystery models, Cicada, Caterpillar, Chrysalis, and Firefly, all seem to be early versions of GPT 5.1. Many users have observed that the Firefly model in particular has been performing exceptionally well in web development. So, this is definitely one of the things you should be excitedly looking out for. Moving on, Pinterest also just announced their Pinterest assistant, which is set to revolutionize the way you shop online. You'll have a visual agent that helps you choose outfits, asks questions about your style preferences, and assists with outfit design. This will be especially helpful for fashion lovers since Pinterest is already widely used to discover and define personal styles. Now, with an in-built assistant, users can easily mix, match, and curate looks directly on the platform. Right now, Pinterest Assistant is available in beta for US users aged 18 and above, and it will gradually roll out to other countries as well. Chat GPT also has a new feature update. Now, when you're using thinking mode and realize you forgot to add some context beforehand, you don't have to wait or restart the entire thinking process. You can now interrupt the process midway and add your context without starting over. This is especially useful when you're doing deep research and are several minutes into the process only to remember a key detail you forgot to include. With this feature, you can simply add it in and continue without having to restart everything from scratch. Moving on, we have news from Anthropic. I know this feature is a bit older, but it's still important enough to mention in this video. Claude now has enhanced capabilities for financial management. This feature eliminates hours of manual busy work, saving you time by bringing Claude's analytical power directly into your documents, spreadsheets, and PDFs. It allows you to create and work with Excel files, making your financial processes much easier. You can now have agents in Excel that go beyond simple analysis, spotting smaller details that might be missed by the human eye. Claude can also identify unique challenges in your financial services. Claude doesn't just work in isolation. It can pull data from traditional banking databases, ERP systems, or even older CSV or PDF reports. All of this is going to transform the way data manipulation works in the long run. This new update allows Claude to create financial reports, assist with audits and compliance, model risks, make forecasts based on data, and generate client-friendly summaries. The world of robotics also saw a huge announcement with the company 1X unveiling its new home robot, Neo. It is a humanoid AI robot designed to handle your household chores for you. It features a friendly, approachable design with a fibercoated top and strong grip, allowing it to assist you with various household tasks. What's even more impressive is its personal memory bank, which lets the robot recall past interactions and contexts it has learned over time. Neo operates autonomously and continues to learn new skills as it goes, making it smarter and more helpful with use. This userfriendly design makes it comfortable to have around the home. If you've ever wanted a humanoid robot assistant, pre-orders are open now with the first deliveries scheduled for 2026 priced at $20,000 for early buyers. Apple also made a huge announcement when it decided to finally fix Siri and pay Google $1 billion annually to use a custom version of Gemini to help power assistant features in Siri. The model is expected to use 1 trillion parameters, far greater than any model Apple has been using. So, these were the major updates with AI that caught our eye, and we're going to be posting one of these episodes every week. So, thanks for watching and I'll see you in the next episode of Context Weekly.