Hi fam!
AI keeps finding new ways to save us time—although it still hasn’t figured out how to attend meetings without creating more meetings.
In this edition, we’re looking at smarter tools, searchable videos, useful shortcuts, and a few AI updates worth knowing before the next five arrive.
The future of AI feels Like Magic, and it’s here!
Ii
In this edition, AI is leaving the chat box and moving into hardware, training, research, and smarter model selection. In other words, it’s no longer just answering questions—it’s choosing tools, coaching employees, analyzing economies, and possibly following you around the living room.
OpenAI's first hardware product, developed with former Apple designer Jony Ive, is reportedly a screenless AI device with built-in speakers that can move around a room and interact naturally with users. Rather than competing with smartphones, it's designed to become a new way of interacting with AI.
Why this matters: The AI race is expanding beyond software. Companies are now competing to define what the next generation of AI hardware will look like.
Alphabet's latest earnings have intensified investor concerns over its AI strategy. Questions around the delayed Gemini 3.5 Pro, rising AI infrastructure costs, and increasing competition from OpenAI and Anthropic have put additional pressure on Google to prove it can maintain its leadership in AI.
Why this matters: The AI race isn't just about launching better models anymore. Investors increasingly want to see clear roadmaps, reliable execution, and sustainable returns on massive AI investments.
Anthropic has integrated its Economic Index directly into Claude. Users can now ask questions about how AI is being used across different industries and occupations, with answers grounded in Anthropic's public dataset.
Why this matters: AI assistants are evolving beyond general knowledge. By connecting them to trusted, domain-specific datasets, companies are turning chatbots into specialized research tools.
Runway has launched Media Router, a tool that automatically chooses the best AI model for generating an image, video, or audio clip. Developers can tell it whether they care most about quality, speed, or cost, and the router selects the model accordingly.
Why this matters: Keeping up with every new creative AI model is becoming a full-time job. Runway is betting that developers will care less about choosing the “best” model themselves—and more about having a system that quietly picks the right one for each task.
Synthesia, known for AI-generated training videos, has launched new features focused on live coaching and interactive learning, moving beyond one-way video creation into real-time employee training.
Why this matters: AI video is evolving from content generation to interactive communication, opening new possibilities for onboarding, education, and corporate training.
Ask AI to Show Its Assumptions
AI can produce a polished answer even when it has quietly guessed half the context.
Before accepting the result, ask:
What assumptions are you making here?
This forces the AI to reveal what it filled in on its own—your audience, budget, priorities, timeline, level of experience, or even what you meant by “better.”
For example, imagine you ask:
Create a marketing plan for my new product.
The AI may assume you have a large budget, an existing audience, and several months before launch. But if you’re actually working alone, spending €300, and launching next week, the entire plan may be useless.
A better follow-up would be:
What assumptions are you making about my budget, audience, timeline, and available resources?
Once those assumptions are visible, you can correct them and get a much more realistic answer.
It’s a small question that helps prevent one of AI’s favorite tricks: turning missing information into confident-sounding fiction.
Turn Any Video Into a Searchable Document
Watching a one-hour video to find one useful sentence is not research. It is hide-and-seek with a progress bar.
A faster approach is to turn the video into a transcript and search the text instead. Tools such as Descript, YouTube transcripts, and NotebookLM can transform meetings, webinars, interviews, lectures, and tutorials into documents you can scan, search, summarize, and question.
For a YouTube video, open the transcript if one is available. For your own recording, upload it to a transcription tool such as Descript. Once the transcript is ready, you can search it like any other document instead of dragging the playhead around and hoping for the best.
Search for words such as: pricing, deadline, customer feedback, next steps…
You can jump directly to the relevant moment without rewatching everything that came before it.
Imagine you missed a 50-minute team call. You probably do not need a summary of every sentence. You need to know what was approved, who is responsible, which deadlines changed, and what still needs a decision.
Upload the transcript to NotebookLM or another AI tool and ask:
List every decision made in this meeting, including the person responsible and any deadline mentioned.
Then follow with:
Which questions were discussed but left unresolved?
That is considerably faster than watching everyone say, “Can you hear me?” again.
The video becomes a reference document you can actually use later.
Transcripts are also useful for editing. In tools such as Descript, clicking a sentence in the transcript takes you directly to that part of the recording, making it easier to find strong opening lines, short social clips, repeated points, awkward pauses, and sections that can be removed.
You are no longer searching the video visually. You are editing by reading.
For anyone who has spent 20 minutes looking for “that one sentence near the end,” this feels suspiciously close to magic.
AI summaries can still miss details, so keep the original transcript and timestamps. When the tool gives you an interesting claim, quote, or decision, check the relevant section before publishing or acting on it.
The goal is not to replace the video with an AI summary.
The goal is to make the video searchable.
Because video is excellent for watching once.
Text is much better when you need to find something again.
From The Archive: Don’t Waste Time Summarizing Meetings. Let AI Do It in 3.1 Seconds.
Back in 2024, we tested Otter.ai, a tool that could turn meetings into searchable transcripts in seconds. At the time, that felt almost futuristic: no frantic note-taking, no replaying an entire call, and no trying to remember who promised to do what.
Today, transcription is only the beginning. Otter can also create summaries, identify action items, answer questions about previous conversations, and turn uploaded audio or video into searchable text.
It is still relevant, but it is no longer alone. Tools such as Fathom, Fireflies, MeetGeek, Granola, and built-in assistants from Zoom and Microsoft now offer similar features, often with their own strengths in integrations, summaries, or team workflows.
What once felt like a clever AI trick has become something we now expect from meeting tools. The real value, however, remains the same: conversations no longer have to disappear when the call ends.
You should not have to attend the same meeting twice—once live, and again while searching through the recording.
Text generation | Image Generation | LMAI recommends |
|---|---|---|
Like Magic AI NFT 🏞️
Our master plan is to publish an NFT image in each newsletter and hand it out to our subscribers. The earlier you subscribe, the smaller the series are. It's a future collectible, a piece of digital art that captures the essence of this moment in time.
Thank you for being a valued subscriber. Together, let's embrace the magic of AI and creativity!
LMAI218-24072026
Was this email forwarded to you? Sign up here 👇
