The CTO must now think like the CFO to survive | Sancus Ventures’ Lake Dai

**Andrew Zigler** (0:06)
Welcome back to Dev Interrupted. I'm your host, Andrew Zigler.

**Ben Lloyd Pearson** (0:10)
And I'm your host, Ben Lloyd Pearson.

**Andrew Zigler** (0:12)
This week, I'm sitting down with Lake Dai, founder and managing partner at Sancus Ventures and a professor at Carnegie Mellon University. She joins the pod to discuss the new operating strategies engineering leaders must adopt in an AI first world. And we dig into why leaders need to think more like CFOs around things like the crushing cost of compute and why AI governance is the biggest blind spot most orgs have right now. You know, we bonded over her role as an educator. I showed that background myself and she's also an advisor. And she explains why the world it seems to be moving so fast, but how it will quickly seem like really old news, which I thought was fascinating. Everyone be sure to slow down and focus on the fundamentals. We really dig into it. And this conversation with Lake really excited for it. But first, we have a roundup of this week's news. So, there are several things that came across our desk, but there was one of the biggest ones that caught my attention, Ben, that I want to talk about first was the phenomenon of engineers talking to their computers these days. We've been seeing a wave of this everywhere. As you know, I've been using voice to text to do most of my agentic coding.

**Ben Lloyd Pearson** (1:21)
Yeah. In fact, you converted me to voice to text just recently. I have also started using it on a daily basis too. It's great.
Yeah.

**Andrew Zigler** (1:31)
No, it saves a lot of time. It removes a lot of friction. I've written about this before even here on Dev Interrupted about how I think the keyword is sometimes the biggest obstacle between myself and building now. So there was a post on that we saw on LinkedIn by Gurgle Oroz of Pragmatic Engineer, talking about how people are doing this in practice, included a really interesting photo. This post went viral, of course, of an engineer literally whispering into this tiny gooseneck microphone coming up right to his face as he worked on something in his IDE. He was standing in a room of folks doing this, and apparently you couldn't hear anybody whispering, all of their agentic coding stuff. So this is fascinating to me, the idea of having these rooms of quietly chanting engineers locked away somewhere. It's a future that Whisper Flow is showing us, a really fascinating dive.
As a voice, the text coder myself, I love to see more people embracing this, and I think more folks should. What do you think about this, Ben?

**Ben Lloyd Pearson** (2:27)
Yeah, I love how you called it chanting. It's like we're invoking software now.

**Andrew Zigler** (2:32)
Yeah, it really doesn't make it seem more spell worthy, speaking of the times.

**Ben Lloyd Pearson** (2:37)
But I've always wondered what it would take to normalize the practice of talking to computers, because the technology has always been there, but it seems like maybe the day has finally arrived. It's always been a cultural problem, not a technical problem. And there's been a lot of companies that have tried it, like Amazon, Google, Microsoft, all the big ones. There's probably a long list of companies that have operated in this space. And there's some that have achieved some success, but they never really take off. They don't become normalized. Like, I can't even remember the last time that I saw someone talk to an Alexa device. Like, no offense if you do and you're listening to this. It used to be so common, and now it's like very rare, it seems like.
But, you know, we're now in the age when the amount of context that you can generate for AI is probably one of the biggest limiters that you personally have for your ability to do things. And speech-to-text is just a wonderful way to quickly download a whole bunch of context because it's much faster than typing. In fact, I wrote almost all of my show notes for this episode using speech-to-text. It was wonderful.

**Andrew Zigler** (3:47)
Yeah, it moves a lot faster. I think like the LLMs, they've come along and they've solved a lot of parts of voice-to-text that before were clunky, like getting the transcriptions really accurate and fast and at your fingertips. But also now we live in a world where you can turn your words into text, but then you can turn that text into tool calls. So you can make a direct link between your voice and making computers do things. That's not a world we lived in even like a year or two ago really.