In this edition we’ll be covering…
OpenAI DevDay's landmark announcements for apps, agents, and Sora 2
The step-by-step breakdown on building AI product ads that convert
A look at how Google's Jules is entering developer terminals
5 trending AI signals
3 AI tools to sharpen your productivity
And much more…
The Latest in AI
OpenAI Dev Day 2025 Unleashes Apps, Agents, and Sora 2
OpenAI just hosted its biggest DevDay yet, and the company's not playing around. With 4 million developers already building on its platform and 800 million weekly ChatGPT users, OpenAI rolled out tools designed to turn developers into AI power users.
Between apps living inside ChatGPT, a full agent-building toolkit, and Sora 2 going API-wide, Sam Altman's crew just made it a whole lot easier to build, ship, and scale AI products.
What's cooking?
Apps in ChatGPT let users chat with third-party apps like Canva, Spotify, and Zillow directly inside the interface. The new Apps SDK (built on Model Context Protocol) means developers can build actual apps instead of basic GPTs. First partners include Booking.com, Coursera, Figma, and Expedia.
AgentKit bundles everything developers need to build production-ready agents: Agent Builder (a visual workflow designer), ChatKit (for embedding chat UIs), expanded Evals (with datasets, trace grading, and prompt optimization), and a Connector Registry for managing data sources across workspaces.
Sora 2 in the API brings OpenAI's video generation model to developers, letting them integrate high-quality video creation directly into apps. GPT-5 Pro and the smaller gpt-realtime-mini model also hit the API alongside a new gpt-image-1-mini for cheaper image generation.
So What?
OpenAI is consolidating the AI stack. Instead of juggling fragmented tools, custom code, and manual workflows, developers can now design agents visually, embed them in apps, evaluate performance automatically, and deploy at scale — all within OpenAI's ecosystem.
Also, with Sora 2 now in the API, we're about to see AI-generated videos flood everything from ads to product demos.
And if you're wondering whether Sam's just flexing for fun, remember: 800 million weekly ChatGPT users means OpenAI's building for scale. The more developers build on its platform, the stickier that ecosystem becomes.
Stay Ahead of The AI Curve
Get bite-sized AI education for just $5.99/mo:
100+ lessons on AI fundamentals, agents, LLMs, prompts & automations
Weekly updates as AI evolves
20+ new lessons released each month
3-Day Free Trial + Cancel anytime
Get Your Hands Dirty!
How to Build AI-Powered Product Ads (That Actually Convert)
With the hype about Sora 2, our team dug in and came up with the industry’s best framework for generating product ads with it.
Forget guessing what makes a good ad. Here's the exact system to research winning patterns, extract what works, and use AI to execute at scale.
Step 1: Research winning ads
Use Tavily's MCP Server in Claude (or Claude Web Search, but it’s a bit less accurate) to find top-performing product ads. Prompt:
Find 10-15 top performing tech product ads from 2024-2025. Include why they worked — specific hooks, pacing, visual patterns, and CTA placement.Real-time search will return breakdowns from Apple, Google, Sony, and other major brands with actual insights on what makes them convert.
Step 2: Extract patterns
Feed the research back to Claude:
Based on these winning ads, identify the 5 most common patterns. What camera movements do they use? How long are the shots? Where does the product appear? What's the pacing? How do they structure the first 3 seconds?Step 3: Generate the prompt
Ask Claude:
Write a 15-second Sora 2 prompt for a SaaS product demo that follows these winning patterns.
Product: [YOUR PRODUCT IN ONE SENTENCE]
Key benefit: [YOUR PRODUCT BENEFIT IN ONE SENTENCE]
Target: [YOUR TARGET AUDIENCE IN ONE SENTENCE]Claude outputs a production-ready prompt following patterns that actually work, not random guesses.
Step 4: Feed to Sora 2
Paste the prompt into Sora 2 along with a screenshot/video reference of your product. First draft should be 80% there. The system works because you're not guessing at what makes a good ad, you're building on proven patterns, then letting AI handle execution.
Tool Spotlight
Google Jules Enters Dev Terminals
Google recently launched Jules Tools, a command-line interface that brings its asynchronous AI coder directly into terminals, CI/CD pipelines, and developer workflows.
As we all know… turns out developers don't want to context-switch to a website every time they need AI help with code.
Jules Tools CLI brings the coding agent directly into your terminal. Spin up tasks, check progress, and validate results without leaving your development environment. Google also made Jules' API public so you can wire it into IDEs, Slack, and custom workflows.
Autonomous task execution sets Jules apart from Google's Gemini CLI. Jules handles scoped tasks independently while Gemini requires hands-on iteration. It notifies you when stuck; otherwise, it runs autonomously until the job's done.
Pricing tiers and GitHub independence launched after Jules exited beta in August: free tier (15 daily tasks), Pro ($19.99/month), and Ultra ($124.99/month). Google's also exploring ways to reduce Jules' GitHub dependency, potentially supporting other version control systems or working without one.
So What?
Google's betting that AI coding agents succeed when they fit seamlessly into existing workflows. Bringing Jules directly into terminals, APIs, and team tools like Slack means developers can delegate coding tasks without leaving their environment.
The bigger play? AI agents are becoming core infrastructure. OpenAI has AgentKit and Codex. Google has Jules. Microsoft backs GitHub Copilot. The race isn't just about who builds the best AI coder, it's about who owns the developer workflow. And whoever wins that battle owns the future of software development.
Speaking of developer workflows, we are cooking with them on Digestibly. Learn to vibe with AI (the right way) on our platform.
Quick Bites
Stay updated with our favorite highlights, dive in for a full flavor of the coverage!
MrBeast warned that AI could threaten creators' livelihoods, calling it "scary times" for the industry as tools like OpenAI's Sora 2 app surge in popularity.
OpenAI and AMD signed a chip supply partnership worth 6 gigawatts of computing power for AI infrastructure, with OpenAI getting the option to buy up to a 10% stake in AMD.
OpenAI acquired AI finance startup ROI, though details on the deal remain scarce as both companies declined to comment on the acquisition.
MIT researchers used AI to map how a new antibiotic targets gut bacteria, accelerating a process that normally takes years down to just months.
Jeff Bezos said AI is "in an industrial bubble" but believes society will benefit despite the current hype cycle around artificial intelligence.
Trending Tools
⌨️ Google Jules - Your async AI coding agent, now available in the command line so you can build, test, and debug without leaving your terminal.
🏗️ Solid - Build full-stack web apps (Node.js, React, TypeScript) with production-grade code you can actually scale and maintain, not placeholder fluff.
📊 Averi - A full-stack AI marketing platform combining powerful AI agents with expert human oversight for campaigns that actually convert.
The Neural Network
Surprise surprise. More compute loading for OpenAI...
Until we Type Again…
Thank you for reading yet another edition of Digestibly!
Reyhan
Early AI leader, advises Fortune 500 companies on AI Development, LLMOps, AI strategy, speaks (for fun) on practical AI. Turns cutting-edge theory into workflows teams can ship today.
Kevin
Ex-AWS SageMaker/Bedrock lead. Shipped infra powering 1M+ devices and $2.5M ARR. Obsessed with high-throughput, low-latency systems— and brings that discipline to every Digestibly release.