No‑BS AI Briefing
No‑BS AI Briefing is for builders who don’t have time for hype. Each episode focuses on a handful of high‑signal stories in AI and AGI, unpacked in simple language with a builder’s perspective. You’ll hear what changed, why it matters, and how you can experiment with the tools, ideas, or strategies yourself—whether you’re leading a team, shipping a startup, or exploring AI side projects.
No‑BS AI Briefing
OpenAI's Super App Bet: Sora Halted, Stargate Scaled & AI Builds AI
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
OpenAI just slammed the brakes on Sora and reigned in its massive Stargate data center plan. A huge strategic pivot that tells us a lot about where AI is actually headed. Meanwhile, Anthropics Claude is writing 80% of its own production code, and Google DeepMind is pouring $75 million into AI tools for filmmakers. We'll unpack what these shifts mean for you, the builder, right now. No BS AI briefing brought to you by Proactive AI. Welcome back, I'm your host, Vikash, and this is where builders get straightforward AI news without the fluff. Alright, let's dive into some high signal items that hit my radar this week. First up, a fascinating partnership from Google. Google DeepMind and A24 are partnering on AI filmmaking tools. The source here is the Google DeepMind blog from June 22nd. What happened is DeepMind, Google's big AI research arm, has launched a first of its kind research partnership with A24, that indie film studio known for really unique high-quality films like Everything Everywhere All at Once and Hereditary. DeepMind is actually backing this collaboration with a substantial $75 million investment. The goal? To co-develop AI tools designed specifically for filmmakers, aiming to enhance their creative workflows and ultimately improve storytelling. Insights they gain from this deep collaboration aren't just staying within the partnership either, they'll feed back into Google's broader AI ecosystem. For builders like us, this matters because it clearly signals a shift toward highly domain-specific creator-driven tools. We're moving beyond general-purpose APIs to AI that really embeds itself deeply into professional workflows. It also sets a pretty strong template for how other creative industries might approach AI, thinking of AI as a co-creator, not just a service you call. And look, $75 million from DeepMind for a partnership like this shows a serious capital commitment to professional great creative markets. It's not just a small grant, it's a big bet. Next, we've got a massive shift from OpenAI. They're reportedly pivoting to a super app strategy, halting development on their Sora video model and scaling back the ambitious Stargate data center project. From what we're hearing, OpenAI is restructuring ChatGPT, aiming to transform it into a more unified, persistent, and agent-based super app that integrates Chat GPT, codecs, and possibly other capabilities. And in a really surprising move, they've apparently halted development on Sora, their much hyped video generation model, and are scaling back those massive Stargate data center projects we've been hearing about. The driving force behind this seems to be a desire to showcase broader utility and a more cohesive product ahead of a potential IPO. So why does this matter for builders? Well, it fundamentally shifts the emphasis from raw, general-purpose models to persistent, task-oriented agents as the primary user interface. If you're building on OpenAI stack, you should expect to see a consolidation around these integrated platforms. You'll need to start designing for agent workflows, not just stateless chat interactions. And this scaling back of Sora and Stargate that could open up significant space for specialists in video generation and infrastructure if OpenAI isn't chasing those hard anymore. Also, some important news in the infrastructure space. Micron and Anthropic are partnering on AI memory, with Micron also making a strategic investment in Anthropic's series H funding round. The new money reported this on June 22nd. The core of the partnership involves Micron and Anthropic co-designing specialized memory and storage subsystems specifically for Anthropic's very demanding AI infrastructure. On top of that, Micron has made a strategic equity investment in Anthropex staggering $65 billion Series H round. And to round it out, Micron is also adopting Claude, Anthropic's flagship AI model, across its own internal operations. As builders, this really underscores how hardware and software co-design is becoming absolutely critical. Performance bottlenecks are increasingly moving to memory and data movement, not just raw compute. This kind of deep integration reinforces the sheer capital intensity required at the frontier of AI development, pushing us to view the entire stack end-to-end from the silicon all the way up to the application. It also hints that tighter collaborative integrations between AI labs and hardware vendors will increasingly shape our future infrastructure choices. Finally, a mind-bending story that feels almost sci-fi. AA Building AI, a report from Business Standard on June 22nd claims that over 80% of Anthropics production code is now written by Claude. This isn't just code generation. The article suggests that AI agents are designing experiments for AI safety research. Systems like Alpha Evolve and something they're calling AI scientists are reportedly generating new algorithms and even drafting academic papers. It means that recursive self-improvement, which has been a theoretical concept, is rapidly moving into practical application within a frontier AI lab. For us builders, this means several things. First, it's going to accelerate iteration cycles to an unprecedented degree. This also means engineering teams are going to have to adopt robust oversight and verification processes for AI-authored code. Your role might shift from writing every line to guiding, validating, and curating what the AI produces. And critically, it alters competitive timelines. If AI can build AI, the rate of capability growth for those who master this could compound incredibly fast. Pretty wild, isn't it? Alright, for our deep dive today, there's one story that really stands out as a potential tectonic shift, and that's OpenAI's reported pivot to a super app strategy paired with the news that they're halting Sora and scaling back the Stargate project. This isn't just a minor product update, it's a fundamental reorientation of their entire business. What happened in simple terms is OpenAI, the company that kicked off this modern AI boom with ChatGPT, is reportedly changing its core focus. Up until now, their trajectory has been very much model first, build the best underlying AI models and offer them via APIs and a somewhat stateless chat interface. Projects like Sora, their incredibly impressive text-to-video model, and Stargate, their plan for a multi-billion dollar hyperscale data center, pointed to an expansion across modalities and a raw infrastructure. But now they're reportedly restructuring Chat GPT into this unified, persistent, agent-based super app. That means less emphasis on just the raw model capabilities and much more on an integrated, remembered experience, almost like a personal AI assistant that lives across your digital life. This pivot is happening as OpenAI is gearing up for a potential IPO and they need to show durable integrated utility beyond just mind-blowing tech demos. Why does this matter right now for us, the builders? Well, it signals a massive platform-centric turn in the AI ecosystem. We're moving from a world where everyone tries to build Elay model to a world where a few dominant players are trying to build the platform. If OpenAI succeeds in creating a sticky, persistent, agent-based super app, it fundamentally changes how users interact with AI and how developers build on top of it. It's less about a single API call and more about integrating into a broader agent-orchestrated workflow. This also has huge implications for data. A persistent agent learns about you and that data becomes incredibly valuable, creating potential lock-in effects. So who should really care about this? Sabor Edge Founders, you absolutely need to pay attention. If your startup is building general-purpose AI applications, you might suddenly find yourself competing directly with OpenAI's integrated offering. Your focus should probably shift even harder towards vertical specialized agents or tools that genuinely augment existing agent workflows rather than trying to build another general chatbot. So product managers. You need to start thinking about agent workflows and user memory. How does your product integrate into a user's persistent AI experience? Is your product designed for stateless interactions or can it leverage and contribute to a user's long-term AI context? Engineering leaders, this impacts your architectural decisions. Do you design for an ecosystem of smaller, composable agents or for integration into a dominant platform? The underlying infrastructure choices matter, especially with OpenAI reportedly pulling back from its own hyperscale data center bets, potentially opening up space for other cloud providers or specialized infra companies, day indie hackers. This could be an opportunity to build highly specialized agent skills or micro agents that plug into these larger platforms, much like individual apps in a mobile app store. You don't need to build the whole OS just a killer feature. How would I think about this as a builder? I'd imagine the future interface not as a command line or a simple chat box, but as a proactive, remembered digital assistant that already knows my preferences, my context, and my goals. Think of it like a smart personal assistant that doesn't forget your previous conversations or preferences the moment you close the tab. For a startup, this means you need to define your niche even more sharply. Don't try to out-super app OpenAI. Instead, identify specific underserved workflows where a deeply specialized agent can add exponential value. For example, a tax agent that specializes only in Indian tax law, leveraging the general AI for foundational reasoning but adding specific domain expert memory and tools. The pullback from Stargate also suggests that while raw compute is important, the value might be accruing higher up the stack at the application and agent orchestration layer. Rhetorical question. Is it time to reevaluate your infrastrategy too? My nobiest take here is this. That might be a signal to other players in that space to accelerate. This isn't just marketing hype, it's a strategic pivot with tangible implications for everyone building an AI. If you want one practical takeaway from today's episode, especially in light of OpenAI's pivot towards persistent agents, here it is. Experiment. Kira prototype a persistent agent for one recurring team workflow in your own team. Here's how to try it in under 60 minutes. One minify a workflow. Pick a simple repetitive internal task that involves text, like drafting internal meeting summaries, triaging support tickets, or generating initial drafts for a specific type of internal communication, e.g. weekly status updates. Two set up an assistant rela, use an existing assistance API from OpenAI, Anthropic, or even a local open source model like Lama with a simple Flask backend. The key is to add a basic database. Even a flat file JSON or a simple CLIT database will do to store conversational history and user-specific preferences. This is your memory. 3. Prompt for persistence and personality. Craft a system prompt for your agent that gives it a personality and instructs it to remember past interactions related to this specific workflow. For example, if it's drafting meeting summaries, tell it to remember key decisions and action items from previous meetings and reference them if relevant. For say test and measure, have your team use this prototype for a week. Track simple metrics like time saved per task or number of follow-up questions needed. Did it actually reduce friction? Did the persistence make it more useful over time? Why is this specific experiment worth your time right now? Because if OpenAI, a leader in foundational models, is pivoting hard to persistent agents, it means they see this as the next battleground for user value by prototyping an agent with memory today. You're not just playing with a new API, you're developing an intuition for how these agentic systems actually deliver value and crucially what the practical challenges are around state management, context windows, and user trust. It helps you design for the future rather than just reacting to it. That's it for today's NoBS AI briefing. If this helped, follow the show in your podcast app and share it with one builder you know. And if you've got questions or topics you want covered, connect with me on LinkedIn and send them over. See you in the next briefing.
