Good morning! Here's what's happening in AI today:
ChatGPT for PowerPoint is now generally available worldwide
GPT-Realtime-2.1-mini adds reasoning and tool use
Gemini Spark can now track topics and react to events in real time
How to vibe code an app the right way
4 new AI tools worth trying today
AI PRODUCTIVITY
OpenAI announced that ChatGPT for PowerPoint is now generally available worldwide across all plans, letting users create and edit presentations directly inside PowerPoint.
ChatGPT can now read uploaded files and use them to update an existing presentation, refreshing summaries, adding new chart data, and creating new slides based on the content it's given, all inside the PowerPoint file itself.
The rollout is worldwide and available across every ChatGPT plan, meaning free and paid users alike can start using it inside PowerPoint immediately without needing an upgrade.
Users can ask ChatGPT to make specific edits such as updating an executive summary, refreshing pricing or forecast data, or adding speaker notes, and the changes apply directly to the deck rather than requiring a copy-paste step.
The feature works as an assistant panel inside PowerPoint itself, meaning the editing happens in the same file the presentation lives in rather than a separate export and import process.
Building a deck has always been one of the most repetitive tasks in professional work: updating the same slides over and over as new data comes in. Bringing ChatGPT directly into PowerPoint and making it free across all plans removes a real recurring task from millions of people's weekly workload.
AI VOICE
OpenAI launched GPT-Realtime-2.1-mini in the API, bringing reasoning and tool use to its Realtime mini lineup at the same cost as the previous GPT-Realtime-mini.
GPT-Realtime-2.1-mini adds reasoning capability to the Realtime mini tier, meaning voice applications built on it can now handle multi-step logic rather than simple conversational exchanges.
Tool use is now supported in the mini tier as well, letting developers connect the Realtime model directly to external functions, APIs, or data sources within a live voice session.
The pricing stays identical to the previous GPT-Realtime-mini, meaning developers get meaningfully more capability without any cost increase for existing or new projects.
The model is available now through the API, meaning developers building voice agents, customer support tools, or real-time assistants can integrate it immediately.
Voice AI has historically forced a trade-off between speed and reasoning depth. Bringing reasoning and tool use into the mini tier at the same price removes that trade-off for a huge range of use cases that don't need the full-sized Realtime model. For developers building voice products, this changes what's possible at the budget tier.
AI AGENT
Google announced that Gemini Spark can now intelligently track topics and react to events in real time, letting users get updates the moment something relevant happens.
Gemini Spark can now monitor ongoing topics such as sports matches, stock movements, or breaking news, and deliver updates as those events unfold rather than requiring the user to check manually.
A sample prompt shown by Google asks Spark to send a match summary and analysis after each U.S. Men's National Soccer match, demonstrating how the feature applies to a specific recurring interest.
The feature extends beyond sports and finance to any topic a user wants monitored, including things like real estate listings or research updates, based on the examples shown alongside the announcement.
Setup happens through a simple prompt describing what to track and how often, with Spark handling the ongoing monitoring and delivery automatically after that.
Most people check for updates by opening an app repeatedly throughout the day. Gemini Spark flips that by delivering the update the moment it matters, removing the constant refreshing that fills a lot of otherwise idle time. For anyone tracking a recurring interest, whether that is a sports team or a market, this turns a manual habit into something that runs on its own.
HOW TO AI
I've watched more people quit vibe coding out of frustration than out of boredom. They open Claude, type "build me an app," and twenty minutes later they're staring at ten random files with no idea which one does what.
That's not an AI problem. That's a thinking problem.
The people actually shipping things with AI aren't writing better prompts. They're doing something less flashy: defining the problem clearly before they ever hit enter. AI writes the code. You still have to bring the thinking. That's the part that compounds.
Here's the framework, stripped down to what actually matters.
Describe it like a human, not a feature list
Skip "build me an expense tracker with categories, charts, and export." Start with the use case instead. Small business owners upload receipts, categorize them, and track monthly spending. One sentence, no jargon. The AI fills in features once it understands what the thing is for.
Tell it who this is for
An app for a founder, a student, and a clinic shouldn't look or behave the same, even if the underlying feature list overlaps. Context isn't a nice-to-have you add later. It's the actual product. I always name the user in my first message now, something like this is for busy freelancers who hate spreadsheets, and the difference in output is immediate.
Ask for the flow before the code
Most people skip this, then wonder why the app feels disjointed. Before any code gets written, ask: Walk me through the app flow. What screens exist, and how does a user move between them? You're not writing code here. You're catching a messy structure before it becomes ten tangled files.
Build one tiny version first
Not the full app. Not the dream version. One input, one action, one output. If it's an expense tracker, that's: enter an expense, calculate the total, show the total. Everything else comes after that one loop actually works.
Make the AI explain the code back to you
Don't just copy-paste and hope. Ask it directly:
“Explain what each file does in simple language.”
“What should I not touch?”
“What will break if I change this?”
You don't need to become a developer overnight. You do need to know what you're shipping.
Debug like a founder, not like a coder
"It's not working" isn't a bug report, it's a cry for help. Give the AI something it can actually use:
“Clicked: Save Expense button
Expected: expense saves and appears in the list
Actual: got a 500 Internal Server Error, nothing saved
File changed: expenses.js, line 42-58”
AI is only as good as the problem you hand it. A vague complaint gets you a vague fix.
None of this requires fancier prompts. It requires explaining what you want clearly enough for the machine to build it. That's the actual shift, and it's the one skill in this whole process that's still yours to own.
Anthropic discovered a global workspace inside Claude's internal neural activations, calling it the J-space, revealing that Claude silently performs reasoning steps like detecting bugs and recognizing staged scenarios before producing any visible output.
xAI added 21 new flagship voices to Grok via API, expanding the range of voice options available to developers building on the platform.
Google Gemini connected Gemini directly to Google Business Profile, letting business owners draft review replies, analyze customer feedback themes, and track revenue trends from inside the Gemini app.
🤖 OpenClaw: Now runs as a local app on Hugging Face so you get a fully local tool-calling agent with no cloud and no API keys required.
🌐 Sakana Translate: Translation tool that handles long text in real time, proofreads tone and phrasing with tracked changes.
🔌 Sarvam MCP Server: Gives your AI coding agents instant gains access to a suite of advanced text and speech tools.
🌍 Copilot in Edge: Copilot can now browse websites on behalf of employees, with IT controlling exactly which pages it can access directly.
THAT’S IT FOR TODAY
Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!
See you tomorrow :)
- Dr. Alvaro Cintas
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial