Was this newsletter forwarded to you? Sign up to get it in your inbox.
I didn’t notice how much typing slowed my thoughts down until I realized I’d stopped.
It hit me on a Saturday afternoon. I was minding my own business when an idea popped into my head that I knew I wanted to capture. I ran to my computer; but instead of my fingers flying to the keyboard, I started talking—and my AI assistant wrote it down.
It happens by muscle memory, as tends to happen when you do something enough times. At this point, I’ve drafted seven full articles this way. “It takes 21 days to form a habit” might be a myth, but in my case, it took around that long for the whole “talking to my computer” thing to take hold.
On the surface, nothing about typing feels laborious—I’ve been doing it since I was a fifth-grade Mavis Beacon all-star. I average 120 words per minute talking; I can hit that while typing if I focus really hard on my fingers. There are tiny frictions that slow me down, micro-decisions about phrasing and punctuation and rhythm that pull focus from where I want my thoughts to be: on the ideas.
These days, I bypass the speed bumps with a setup that’s become like an extension of my brain. First, there’s my Working Overtime workspace, a ChatGPT project I have set up to help me brainstorm ideas, do research, and flesh out drafts for my column. Then there’s Monologue, the AI-powered transcription app that Naveen Naidu Mummana, entrepreneur in residence at Every, is building (it’s currently in public beta). I just press a key to talk; Monologue relays the message; the alien superintelligence inside my project tells me what it “thinks.”
ChatGPT has its own built-in voice-to-text mode, and there are other popular voice dictation tools on the market, like Whispr Flow. But I found frustrations with each: The dictation lags, or the output is unclear, or opening the app slows down my computer to the point where I can’t use it. Using Monologue paired with ChatGPT was the first time that talking to my computer felt like the evangelists promised it would: easy.
Once I stopped treating the keyboard as my only entry point, the whole shape of my work changed: Ideas flow faster, structure emerges in conversation, and clarity comes from rounds and rounds of “How’s this?” and “What about that?” If you haven’t tried talking to your computer yet, I highly recommend it. Let me tell you why.
Talking as a mode of work
Here’s how it looks in practice. I open a fresh chat inside my project and say, “I have an idea for a Working Overtime essay about talking to my computer instead of typing and how it’s changed the ergonomics of my work.” The first few times I did this, my dog perked her ears up from her chair in the corner, thinking I was talking to her. Now, she keeps on snoozing.
The Only Subscription
You Need to
Stay at the
Edge of AI
The essential toolkit for those shaping the future
"This might be the best value you
can get from an AI subscription."
- Jay S.
Join 100,000+ leaders, builders, and innovators
Email address
Already have an account? Sign in
What is included in a subscription?
Daily insights from AI pioneers + early access to powerful AI tools

Comments
Don't have an account? Sign up!
Many of Every's software products seem to be aimed/limited to the Mac. Will Monologue be the same, or is it a web-based service, thus available on Linux workstations?
I sent you a note on X - not seeing this here. anyway - enjoyed the piece. I have a Chat GPT window open constantly called Human IO interface. It is an exploration of how we process information visiually far faster than any other way - but we can get information out faster through speech than any other method available (at least til Elon gets that Neurolink working). seems like an inefficient system - like we should be able to use all that visual processing power to project information - but here we are talking to our computers and listening to podcasts
I'm interested in your thesis and experience here, but it seems a little too entangled with what your specific AI projects and transformative aspects of AI can do. In other words it feels as much about the setups and dialog practices you have with AI as it does about "thinking/talking out loud" and using realtime transcription. Obviously for you, you're already used to what your existing projects helped you do, so the transformation you're experiencing is real, but how translatable is it, and perhaps more precisely, how much of its value is in the *combination* of realtime transcription and well-setup projects/contexts?
I suppose one way to disentangle that a bit might be to be trying it with something a little less formalized or refined in terms of context. It may be a little bit of an academic distinction, but to me it feels like a way of orienting toward where the real gains are. I find the idea of free-form talk and immediate transcription to be a bit messy and imprecise, and the implication is that AI will help make up for that, but how much setup/providing of context does that require? Or does it work well "out of the box" with a good general purpose LLM already? I suppose I can just try it 😄
@Oshyan I think you are right - this project is about execution. My read is that Katie's experience is that typing substantively slows us down, perhaps even derails us. The goal is therefore to not just create greater interface fluidity, but thinking fluidity as well. In the background, she still has all the contextual elements of her work (style sheets, execution sequences etc)