Everyone's Job Has Changed. Forever.
I sent a note to our team earlier today about how, in an AI-native world, your job is not to do the work—it’s to ensure the work is great.
Team,
I’ve been thinking a lot about Tobi’s note that was shared in #tool-ai — I tweeted about this over the summer and it is so aligned with how I see the world. This isn’t a fire drill - AI is not a trend. It’s a permanent shift in how work gets done. And we’ve already been moving in this direction as a company.
We’ve given everyone an AI budget and the space to experiment with agents. We’ve built internal tools that lean on LLMs. We’ve started integrating AI into our development, marketing, onboarding, and community engagement.
But we’re still only scratching the surface. I’ve always believed in giving away our Legos—sharing what we’ve built so others can do better and faster. That applies here too. We should be sharing our agents, prompts, setups, and learnings with each other constantly. What’s working? What’s surprising? Where did AI fail you—and how did you work around it? This is absolutely not just for engineers either. This is for all of us. Product. Community. Ops. Design. Leadership. Everyone. That’s why we’ve made this a core part of how we work—not a side experiment.
But we need to go further.
The next frontier is collective learning. Pushing boundaries. Making agents for each other. Teaching each other. Showing off. Failing in public. Giving away our legos. We are on the edge of entirely new workflows and ways of thinking—and our advantage is speed. Velocity. Taste. Execution. That only compounds when we learn together.
So here’s how I’m thinking about it:
Everyone is QA now. In an AI-native world, your job is not to do the work—it’s to ensure the work is great. Quality, taste, and output are now everyone’s domain, regardless of role. The faster we embrace that shift, the faster we win.
Prototypes should start with AI. Whether you’re making a new internal doc, building a feature, or writing a tweet—start with AI. Make it good with AI, then make it great with your craft. Show don’t tell has an entirely new meaning and every single person who works at Ephemera is a builder. A picture was worth a 1000 words. A mock-up was worth a 1000 pictures. A demo was worth a 1000 mocks. And now a proto-type is worth 1000 demos. The best way to build the future is to feel it. Make things, share it, use it, break it, improve it. That’s how you build great things.
No headcount asks without an AI plan. Before we add more people, we have to ask: How far can we get with agents, automation, and smarter workflows? We’ve done it before and seen wild results. This is for real. Show off your workflow and how you’re using AI to 10x or 100x your output before thinking about “we need to hire more people.” We will obviously still hire, but we should obsess on our systems first.
Make it visible. Show your work, share your tools, and brag about your agents. Let’s fill Slack with experiments, prompts, and weird wins.
The fastest way to lose with AI is by trying to protect your job because it protects your ego. Let it go. The job is output, not work anymore. Train your team; they aren't just people anymore.
This is a massive unlock. But only if we push it. I’m all in—and I hope you are too. If you’re unsure where to start, ask someone deep in it. If you are deep in it, help bring someone else along. Let’s do the best possible job of figuring out our craft in an AI-native world together.
—Shane
I deeply believe this: In an AI-native world, your job is not to do the work—it’s to ensure the work is great. Its human job is about orchestration and coordination - and being responsible for the quality of the output.
It’s almost as if every single human on earth just became a manager. We all now have tons of different team members who all work on their teams to get things done and we have to ensure quality. The only difference is that your team is now a bunch of agents. They are specialized, infinitive learners who change daily, are constantly updating, and are even doing what I did yesterday better than I did it. This can be hard to grasp but the job security of the future isn’t in the person who can do the task the fastest, this will be won by AI. It’s in the person who can manage the agents, connect all the dots between them, and bring the work and our teams of people together to know what matters most, what our bar for quality is, and ensure that people want what we are building.
What’s funny is what hasn’t changed at all: building something people want. That’s all that matters, and it’s what has always mattered. People who use things don’t care how things are built; they care about things they love to use.
Build something people want. More true today than ever. We can just go 100x faster with a smaller team. That’s the secret hidden in plain sight, today.
Update: A member of our team posted this when I shared this with our team
AI is NOT a replacement for being the expert in your area. I can't stand when people share docs with me that are clearly AI written and expect me to be the human touch to make sure what it is saying is accurate and clean it up so that it doesn't look like a robot wrote it. AI is not an excuse to do more and be lazy and push the work on to the other people to read your AI garbage. You still need to read and edit your work even if AI wrote the first draft.
Same goes for code. I vibe-coded the hell out of my bot last week because I didn't have time. The code is not great. It's not optimized. I would never put my name on it and tell people this is exactly how you should build this bot. Pushing up PR written by AI and putting the work on other engineers to code review and give feedback on your robot implementation is also not an excuse. I don't care if you vibe code it. But don't put the work on other people to take it from a it works implementation to a great one.
And here's my response because I think she hit the nail on the head:
She nailed it. This is the most important thing to understand with AI — and something I see too often getting missed:
You are now the reviewer.
If you’re using AI, you are responsible for the output — docs, code, designs, content, whatever. If it’s junk, that’s on you, not the AI.
The goal isn’t more output. The goal is better output.
That means editing, shaping, clarifying, and doing the real work to make something great — not just accepting whatever AI gives you.
Ask yourself: Would someone actually want this? If not, fix it. The real skill with AI is constraint and craft.
Draw tighter boxes. Edit 100x more than you think. Like Twain said, “If I had more time, I would have written a shorter letter.” - And now you have more time, and the time you have is to make things great.
It takes work to get great output. But that’s the job now.
Use AI a lot. Share what works. Help each other raise the bar. More isn’t better. Better is better.