Everyone Is Obsessing Over
The Wrong AI Tool.

Key Takeaways

Most people obsess over which AI model is best. John argues the bigger day-to-day bottleneck is simpler: how clearly and quickly you can tell the AI what you want.

The main idea:

• AI work is mostly intent. Coding agents, chatbots, writing tools, image generators, research tools, and automation systems all perform better when your instructions are detailed and clear.
• Typing creates friction. When typing feels annoying, you tend to send shorter, vaguer, half-formed prompts. That means the AI gets weaker instructions.
• Typeless turns speech into usable text. You speak naturally, and Typeless removes filler words, cleans up repetition, formats the result, and turns rambling into polished text.
• It works across the tools you already use. John mentions Google Docs, Notion, Gmail, social posts, Claude prompts, video planning, and general AI workflows.
• The Ask Anything feature reduces app-hopping. Instead of asking ChatGPT for an email, copying it, pasting it into Gmail, and editing it, you can ask Typeless to generate the message directly inside the workflow.
• It is more than dictation. Old voice-to-text often created cleanup work. Typeless is useful because it cleans up the output as part of the process.

John's challenge: try dictating your longer AI prompts, emails, and messages for a week. Then ask whether your AI got better instructions because you spoke instead of typed.

Frequently Asked Questions

Full Video Transcript

Transcript of "Everyone Is Obsessing Over The Wrong AI Tool" by John Elder

The Real AI Bottleneck Jump to video

My number one AI tool right now isn't Claude, ChatGPT, or OpenClaw. It's a voice-to-text app called Typeless.

Because the more I use AI, the more obvious this gets. The bottleneck isn't the model. The bottleneck is how clearly I can tell the model what I want. And typing is a terrible way to do that.

What Typeless Does Jump to video

Typeless is an AI voice dictation app. You talk naturally and it turns what you say into polished text you can use in any app. Emails, documents, prompts, notes, tweets, messages, whatever. The big promise is simple: speak, don't type.

Sounds like a small productivity hack, but once you use AI all day, you realize something. Modern AI work is mostly explaining your intent. Claude, ChatGPT, coding agents, image generation, research, email automation, all of it works better when your instructions are clear. And Typeless makes that part faster.

Why The Input Layer Matters Jump to video

Most people talk about AI like the only thing that matters is the model. Claude versus ChatGPT, GPT versus Gemini, which coding agent is better, which app has the smartest brain. But there's another layer in the stack that almost nobody talks about: the input layer.

If your prompts are short, vague, or half-formed because typing them is annoying, it doesn't matter how smart the AI is. You're feeding it weak instructions.

That's where Typeless changed things for me. I can speak a long prompt a lot faster than I can type it. And Typeless removes filler words, cleans up repetition, formats the results, and turns my rambling into something I can actually use.

So instead of thinking, "How little can I type to get this done?" I can explain what I actually want. It's a huge shift. And once you understand this input layer, you start seeing it everywhere.

Long AI Prompts Jump to video

Let me show you the most obvious use case first: a long AI prompt. Normally, if I want to give Claude a detailed task, I have to type something like, "Look at this project, figure out what files matter, summarize the structure, tell me what's missing, and give me the next three steps."

But in real life, a good prompt is much longer than that. I want context, constraints, what not to do, and how I want the answer formatted. Typing that out is all friction.

With Typeless, I just say it. Something like, "Look at this YouTube project and turn it into a clean to-upload checklist. Keep the sponsor requirements separate from the creative notes, flag anything that still needs a link, screen recording, or final approval, and then give me the exact next actions in order."

That's the magic. I didn't write a careful prompt. I talked through what I wanted and Typeless turned that into something usable. This matters because across all AI tools, the main bottleneck is still how clearly you can express your intent.

Ask Anything For Email And Posts Jump to video

Now you're probably thinking, "Okay, isn't that just dictation?" And that's what I thought, too. The next part is where it stops being just speech to text.

Here's a different kind of use case: email. I don't want a workflow where I have to go to ChatGPT, ask it to write me an email, copy the results, paste it into Gmail, edit it, and then send it. That's too many steps.

Typeless has an Ask Anything feature, so you can ask it to generate something directly from inside the Typeless workflow. For example: write a short email rescheduling my meeting with April from this afternoon until tomorrow morning. Keep it friendly, apologize briefly, and ask if 10 a.m. works.

Now I have that message without bouncing between three different apps. That's the difference between using AI as a separate destination and using AI as part of how you operate your computer.

Same thing with a tweet: generate a tweet about how AI productivity is not just about better models, it's about removing input friction. Or a LinkedIn post: draft a short LinkedIn post introducing a tool that lets people speak instead of typing. This is where Typeless starts feeling less like dictation and more like an AI command layer.

Works Across Everyday Tools Jump to video

But the email demo is only one surface. The real test is whether it follows you across different tools you already use, and Typeless does.

I can use it in Google Docs, Notion, Gmail, social media posts, Claude prompts, or while planning a video. Really anywhere.

And the output isn't just raw speech to text. It removes the ums and ahs, and also the repeated phrases. If I change my mind mid-sentence, it cleans that up. It formats the results so it reads like something I typed on purpose.

That's the part normal dictation never got right for me. Old voice-to-text felt like I was creating cleanup work for myself. Typeless feels like I'm skipping that cleanup work. If you've tried voice-to-text before and didn't like it, that's probably why. The old version made you clean up the mess. This one cleans it up for you.

Typeless Plus AI Agents Jump to video

Here's the advanced version of this, and this is where it connects back to the AI tools I use every single day.

Earlier this week, my desktop was an absolute mess. Random PDF files, screenshots, project files, app icons, the normal chaos. I opened Claude Cowork and used Typeless to speak a command into it. Something like, "Look at my desktop, group the files by projects, create folders where it makes sense, archive anything old, and then tell me what's left."

Cowork handled the desktop actions. Typeless handled the input. That's an important distinction. Typeless isn't the agent moving files around. Typeless is the layer that lets me give the agent a clear command without typing the whole thing out.

That's why I keep calling it the input layer. Coding agents, general LLMs, writing tools, email, research, image generation, automation systems, they all need the same thing from you: clear instructions. Typeless makes clear instructions easier to produce. That's why this is bigger than, "I found a faster way to type."

Why Typeless Stuck Jump to video

I've tried other voice tools out there. Whisper Flow has a big following, and a lot of people like it. For me, it felt a little finicky. Small errors stacked up across the day and drove me crazy.

Typeless is the one that stuck because it just feels simple. I speak naturally, it cleans up what I said, and I can use it where I already work. That matters more to me than having another separate AI app to manage.

In full disclosure, this video is sponsored by Typeless, but I used their service long before they contacted me. The reason I said yes to the sponsorship is because this is exactly the kind of tool I think AI people should be paying attention to.

Trial Offer And Final Test Jump to video

Everyone is trying to add more AI to their workflow: better models, better agents, better tools. But if you're still typing everything into those tools, you're leaving a huge bottleneck in place. And that bottleneck is the input layer.

Typeless is available for Mac, Windows, iOS, and Android. If you want to try it, use the link in the description below to get a one-month free trial plus $5 in credits after the trial ends.

Try dictating all of your AI prompts instead of typing them for just a week. Not everything, just the long prompts, maybe emails, messages, and other stuff where typing slows you down.

At the end of the week, ask yourself one question. Did you give your AI better instructions because you were speaking instead of typing? For me, the answer was yes. The keyboard isn't dead everywhere, but for AI work, it's no longer the center of my workflow. Check it out today.

About John Elder

John Elder has been coding for over 30 years. He runs Codemy.com, an online coding education platform where he's taught over 20 million students, and a YouTube channel with over 250,000 subscribers. He also runs JohnElder.AI, where he teaches AI, Python, and agentic workflows.

John is based in Las Vegas, Nevada and has authored multiple courses on Python, Django, Tkinter, and AI development. His teaching philosophy focuses on practical, real-world coding skills, not theory.