This website uses cookies

Read our Privacy policy and Terms of use for more information.

Hi fam!

AI apparently had a busy few days while the rest of us were trying to survive August. 😅

This time, we’re looking at AI that can turn Reddit posts into videos, take over boring workflows, understand our voices better, and make agents easier to control. Plus, we’ve got a practical tool worth trying, a simple AI trick, and another useful find from the archive.

The future of AI feels Like Magic, and it’s here!

AI is getting more visual, more vocal, and much better at doing things on its own. From AI-generated Reddit videos to smarter agents and booming creative tools, here are the updates that caught our attention.

  • Reddit is testing AI-generated videos and audio versions of posts and comments. Selected Reddit conversations are being transformed into short videos narrated by AI voices, with the relevant text highlighted as it is spoken. Reddit says the experiment is currently limited and the posts are manually selected.
    Why this matters: Reddit is sitting on an enormous archive of human conversations. Turning that text into automatically generated video and audio content could give old posts an entirely new life—and shows how quickly the line between written and video platforms is disappearing.

  • AI automation startup Relay is shutting down, with members of its team joining Google’s Chrome division. Relay launched as an AI-powered workflow automation platform and had positioned itself as a potential alternative to tools like Zapier.
    Why this matters: The interesting part isn't only that another AI startup is disappearing. Its team is heading to Chrome, just as browsers are increasingly becoming places where AI agents can actually perform tasks for users. The browser itself may become one of the biggest battlegrounds for AI automation.

  • Wispr Flow has reached a $2 billion valuation after raising $280 million—and launched its own speech-recognition model. The voice-to-text company says its new Canto model significantly reduces transcription errors in challenging real-world environments. Wispr Flow is already used by more than 10,000 businesses.
    Why this matters: Typing may slowly be losing its monopoly on how we interact with computers. Investors are betting heavily that talking to our devices could become a normal way to write, search, create, and work—not just something we do with voice assistants.

  • Higgsfield has raised $400 million, pushing the AI video startup’s valuation to $5.4 billion. Its valuation has more than quadrupled in roughly six months as demand grows for AI tools capable of producing advertising and social-media video content.
    Why this matters: AI video is rapidly moving beyond “look what this model can generate.” Investors are betting that these tools will become part of real commercial content-production workflows, particularly advertising and social media.

  • GitHub is experimenting with “canvases” to make AI agent workflows easier to see and control. Instead of letting long-running agent work disappear into an endless chat thread, the approach puts tasks, outputs and decisions into a more visual workspace where users can inspect and steer what the agent is doing.
    Why this matters: As AI starts handling longer and more complicated jobs, chat may simply be the wrong interface. The next generation of AI tools may look less like conversations and more like visual workspaces where humans supervise multiple AI tasks.

Turn Feedback Into Rules

You know that moment when you've corrected AI for the fifth time?

Shorter.
Less corporate.
Don't use bullet points.
Stop saying “game-changer.”
And please, for the love of productivity, stop ending every answer with three more questions.

Instead of repeating the same corrections forever, make AI turn your feedback into instructions.

After you've worked together on something for a while, ask:

Based only on the corrections and feedback I've given you in this conversation, create a short set of rules for how you should handle this type of task next time. Don't add preferences I haven't explicitly expressed.

AI will turn all those scattered “not like that” moments into a reusable set of instructions.

You can then save the rules and paste them at the beginning of your next conversation—or add them to your project instructions if the tool you're using supports them.

For example, after editing a few newsletter articles, you might end up with:

Keep paragraphs short. Use humor sparingly. Avoid corporate language. Don't overuse lists. Keep headlines under 60 characters. End with a clear takeaway, not a follow-up question.

The trick works especially well for writing, recurring reports, social posts, research formats, client work, and anything you regularly ask AI to do in the same style.

You're already training AI every time you say “No, more like this.”

You might as well keep the instruction manual.

Make a Tutorial Without Writing One

Someone asks, “Can you quickly write down how to do this?”

And somehow “quickly” turns into screenshots, cropping, arrows, numbered steps, and 30 minutes of explaining something that took five minutes to actually do.

Scribe flips that process around.

Start recording, perform the task normally, and Scribe automatically captures your clicks and turns them into a step-by-step visual guide with screenshots and instructions.

Need to show someone how to publish an article in WordPress? Start Scribe, publish one article as usual, stop recording—and your first version of the tutorial is already there. Clean up any unnecessary steps, add a little context, and share the link.

It works especially well for onboarding, website updates, internal software, recurring admin tasks, or anything you've already explained more than twice.

Instead of:

Do the task → Take screenshots → Write instructions → Format tutorial

you get:

Do the task → Tutorial exists.

From the Archive: Turn Any Video Into a Searchable Document

Watching a 45-minute video again just to find that one thing someone said somewhere around minute… 27? Not exactly efficient.

In this archive pick, we showed how to turn a video into a searchable document, making it much easier to find specific information, pull out key points, and reuse what’s buried inside.

Today, Scribe turns what you do on screen into documentation. This one tackles the opposite problem: turning what you watch on screen into information you can actually search.

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!

LMAI225-18082026

Was this email forwarded to you? Sign up here 👇

Reply

Avatar

or to participate