In partnership with
October 16th, 2025
Happy Thursday!
Welcome to another issue of How To AI. We’re glad you made it.
Here’s what we’ve got for ya:
🎤 Turn Slides into Talking Points with AI
🔒 OpenAI Tightens Safety Controls with New Routing System
🌐 Huawei Commits to Open-Source AI Stack by December
🚁 DJI Loses Legal Battle Over Military Company Classification
🗣️ Quick Hits: Product Upgrades, Policy & Governance, Content & Discovery
Let’s get started!
The Future of the Content Economy
beehiiv started with newsletters. Now, they’re reimagining the entire content economy.
On November 13, beehiiv’s biggest updates ever are dropping at the Winter Release Event.
For the people shaping the next generation of content, community, and media, this is an event you won’t want to miss.
🎤 Turn Slides Into Talking Points With AI
Stop Reading Bullets, Start Speaking With Confidence
Most presentations fail because the speaker just reads the slides. The audience tunes out, and the message gets lost. AI changes that. By uploading your slide text into ChatGPT (or whatever AI you choose), you can get back clear talking points or a short narrative that guides delivery. Instead of memorizing lines, you’ll have a natural flow that keeps attention where it belongs: on you.
Step 1: Upload your slide deck to ChatGPT
It’s simple to upload any presentation (or any kind of file) to most all of the GPT tools and they all work similarly. Next to the box where you would “chat”, click the + to upload a file. Select your PowerPoint file and move on to step 2. Alternatively, if you are using Google Slides, you can go to File → Download and download the deck in either PowerPoint format (PPTX) or PDF, and then upload it.
Step 2: Ask AI to draft your talking points
In the same chat with the slide deck upload, copy and paste in this prompt-
You are my presentation coach. Turn the following slide text into speaker notes that sound natural when spoken aloud.
For each slide, give me 2–3 talking points in plain English, no jargon, max 25 words each.
Keep the flow conversational and audience-focused.
Format like this:
Slide 1:
- Talking point
- Talking point
Slide 2:
- Talking point
- Talking pointThe model will return short, easy-to-say notes instead of dense paragraphs. These become your guide for delivery.
Step 3: Refine for clarity and tone
If the first draft feels stiff or too generic, ask for adjustments:
Make the talking points more persuasive for an executive audience. Keep them concise and confident.
Or:
Rewrite the notes to sound like I’m explaining this to a client who doesn’t know the technical details.
This lets you tune the voice to match your audience without rewriting everything yourself.
Step 4: Build a narrative arc (optional)
If you want more than slide-by-slide notes, ask:
Create a short opening statement, a transition sentence between each section, and a closing summary based on the slides.
This gives you a full storyline to carry the room from start to finish.
Step 5: Practice with the notes, not the slides
Print or save the AI-generated talking points as your speaker notes. Rehearse from them instead of staring at the deck. You’ll sound more natural, and your slides will support you instead of competing with you.
📌 Key takeaway: AI turns static slide text into dynamic talking points, helping you present with clarity and confidence instead of reading bullets off the screen.
The Rundown: OpenAI deployed a new safety routing system that automatically switches ChatGPT to GPT-5 during emotionally sensitive conversations, and introduced parental controls for teen accounts.
Details:
The router detects risky conversations mid-chat and temporarily shifts to GPT-5, which uses “safe completions” to handle sensitive queries.
Parental controls allow settings like quiet hours, disabling voice mode and memory, and alerts if the system detects potential self-harm.
The changes follow criticism of GPT-4o’s overly agreeable nature and a wrongful death lawsuit tied to a teenager’s suicide.
Takeaway: OpenAI is moving from speed to safety, a shift that will ripple across customer-facing AI. If you deploy chatbots in sensitive contexts, run a safety audit now: test escalation paths, review parental controls, and set clear guardrails before regulators or lawsuits force the issue.
The Rundown: Huawei announced it will open-source its full AI software stack by December 31, 2025, including the CANN toolkit, Mind series development tools, and openPangu foundation models.
Details:
CANN will expose compiler and virtual instruction set interfaces, while other components remain partially proprietary.
The Mind series toolkits and SDKs will be fully open-sourced, giving developers control over debugging and libraries.
Huawei pledged PyTorch and vLLM compatibility, plus modular OS integration with Linux distributions.
Takeaway: Huawei is betting that transparency will win developer adoption and challenge NVIDIA’s closed ecosystem. If you rely on AI infrastructure, mark December as a checkpoint: review Huawei’s release quality, licensing terms, and community traction before deciding whether to invest engineering time.
The Rundown: A federal judge rejected DJI’s lawsuit challenging its placement on the Department of Defense list of Chinese military companies, citing evidence of contributions to China’s defense base.
Details:
Judge Paul Friedman pointed to modified DJI drones used in the Russia-Ukraine conflict as proof of military applications.
DJI argued it only makes consumer and commercial drones and said the listing caused financial and reputational harm.
The company faces a potential US sales ban in December unless agencies rule its products pose no national security risk.
Takeaway: The ruling shows how dual-use tech can trigger restrictions based on potential applications, not stated intent. If your operations depend on DJI hardware, prepare contingency sourcing now, since a December ban could disrupt both commercial and government contracts.
Product Upgrades 🚀
📱 Snapchat limits free Memories storage to 5GB, adding paid tiers from $1.99 monthly to offset costs for users with thousands of saved Snaps.
🎬 YouTube Labs launched AI music hosts that share stories and trivia during listening sessions, with early testing limited to US participants.
📊 OpenAI rolled out ChatGPT Pulse for $200 Pro subscribers, generating personalized morning briefs overnight to encourage daily use.
Policy & Governance 🏛️
🇺🇳 UN launched the Global Dialogue on AI Governance, giving all 193 member states representation in shaping international AI oversight.
💻 Microsoft added Office Agent, powered by Anthropic models, enabling Word and PowerPoint creation while expanding beyond OpenAI.
🇨🇦 Canada launches national AI Strategy Task Force, beginning a 30‑day public consultation to shape its next national framework.
Content & Discovery 📚
🎨 Adobe released Firefly Boards, an AI moodboarding tool that mixes models from Adobe, OpenAI, Google, and Runway on a shared canvas.
🏥 MIT developed MultiverSeg, which learns medical image segmentation from user clicks, cutting annotation time for researchers from days to hours.
🎥 Meta debuts “Vibes”, an AI video feed inside its app for remixing and sharing short clips, integrated with Instagram and Facebook.
🎥 Smarter PowerPoints with AI
This demo shows how to use AI features inside PowerPoint to instantly redesign slides, generate timelines, add live video, and rehearse with real-time coaching. You’ll see how to turn dense text into clear bullets, practice delivery with automated feedback, and add subtitles for accessibility. It is a fast way to make your slides look professional and your presentation sound confident without extra effort or outside tools.
📫 Ask the Inbox
Q: “If I let AI write my speaker notes, won’t I sound robotic or scripted?”
A: Not if you set the right boundaries. AI can draft the structure, but you control the delivery.
Start with raw text: Give AI only your slide text, not a blank prompt. That way, it reflects your content, not generic filler.
Ask for bullets, not paragraphs: Prompts like this work well:
Turn this slide text into 2–3 short talking points per slide.
Keep them under 25 words each, plain English, conversational tone, audience-focused.
<insert slide text here>Tune the tone later: Run a second pass if needed (“make this sound like I’m explaining to a client” or “sharper for executives”).
Practice with edits: Read the AI notes aloud, then rewrite phrases that don’t sound like you. Keep only what feels comfortable in your voice.
📌 Key takeaway: Use AI as scaffolding, not a script. Let it break down your slides into clear beats, then adjust so your personality carries the room.
🤔 Got Questions?
Have a question you want answered? Email [email protected] and you just might be featured in an upcoming issue!
This newsletter is your starter kit for mastering AI with confidence. We keep things simple, show real examples, and focus on quick wins you can repeat.
Stay curious, stick with it, and watch your skills grow week by week!
Until next time :)