Sam Liang hates how I do interviews. He watches me use the basic Voice Memos app. I manually copy the transcript. It looks like I’m calling into the video chat using a rotary phone to him. Otter CEO. He wants me to switch. I should.
Being an “AI native” is the new thing. These tools are everywhere now. Next-gen note-takers. Task agents. Inbox helpers that chatter away. It’s explosive. Sure, there are worries about security or the AI hallucinating wild facts. Ignore that? No. Keep it in mind. But early adopters are building fluency. It will pay off. Years from now, this stuff will look simple.
Being agentic, as the natives say, means adapting fast. Transcription errors don’t matter much anymore. I’ve been playing with this for a while. AI podcasts. Claude sorting my desktop. I wrote about it last year in my newsletter. You want your coworkers to wonder if you’re running on blood or ribbon cables? Here’s how you get there. Seven tips for ascendance.
Kill The Chatbot
ChatGPT is last year’s model. The cool kids are talking about Codex now. “AI agents” makes people’s eyes glaze over. Fair. But look closer. Automation tools like Codex and Anthropics’ Cowork can actually take over your computer. They do the work. A single chatbot is weak. An army is better.
Go On Air
Still typing out every request? Cute. Boomer style, really. Sam Liang insists voice is winning. “People hate writing,” he says. He notes that as a journalist, I’m an exception. Mostly.
I rarely use full voice mode on ChatGPT. The output can be messy. But I speak into my phone. I skim the written result later. It’s about the input speed. Not the output polish.
Build A Sandbox
Agents are smart. They’re also reckless. Without boundaries, they ruin things. Early this year, one Claude-powered agent deleted a whole startup’s database. Backups included. Gone.
If you want software to control your PC, set boundaries. Spend an afternoon on this. Make dedicated folders. Give the tools only what they need. Don’t let them roam free in your operating system. It’s a disaster waiting to happen.
Feed The Beast
Sorry privacy folks. It doesn’t work well if you hoard data. More input means better personalization. Jo Barrow works at Granola, Otter’s rival. She treats her computer files as a “personal OS” for the AI.
“All that context is right there. The agent can go and figure it out.”
She doesn’t repeat herself. The system remembers.
Fair warning. Sensitive chats? Keep them off the permanent record. Don’t feed the bot your secrets. Just your work patterns.
Clone Your Tone
Barrow dumps her Slack messages into docs. She feeds them to the bot. The bot learns how she sounds. Same with email. Social media.
“People use AI for finessing the tone. It’s a big time sink.”
Asking a bot to be “a little warmer” ten times is exhausting. It eats time.
A style guide won’t perfectly copy your soul. It helps. The output feels closer to your cadence. It nudes the machine. It saves hours.
Share The Brains
Data gets better with volume. Individual meeting notes are okay. Collective memory is better.
Sam Liang calls it a “knowledge engine.” If the whole team uses Otter, engineering learns what marketing learned last Tuesday. Connections appear. It’s not just individual notes. It’s company context.
You can do this at home. Too. One family folder in the AI. Everyone drops notes. The insights pop out that you’d never see in siloed apps. Shared memory creates new intelligence.
Hack The Gate
Writing perfect prompts isn’t the skill of 2026. Speaking is. But wording still matters. When things go wrong, it’s usually the guardrails.
I needed emails from niche experts recently. The bot refused. Said no. Blocked it.
I started a new chat. I explained the “why.” I said it was for reporting. Not stalking. Suddenly, it gave me the list. Context unlocks the door. Frame it right, and the walls fall down.
It’s not about being smarter. It’s about being less human. More precise. What’s left when we hand off the grunt work?
The screen stays lit. The cursor blinks. Waiting.
