Careers
Agent-Native Internship
Here's the truth: we have no idea how this whole AI thing plays out. (No one does). However, a few entrepreneurial friends have somehow found ourselves at the bleeding edge of agentic software development.
We got ChatGPT immediately when it came out. Then Claude. Then Cursor. Then OpenClaw. Then Hermes. Now Buzz. (And everything else obviously).
We built software, systems and tools because the technology was constantly getting better, but then things started getting wild in Jan/Feb 2026.
We were building custom harnesses to make agents turn raw intelligence into predictable behavior that actually started running our companies.
And then during one late night conversation, I happened to tell my friend that I hadn't really seen anyone else who's figured out how to do what we're doing right now.
Oops. Turns out almost no one on Earth knows this yet. (We checked, and we know all the same YouTubers).
We've gone to the in-person AI engineering events and realized we're ahead of (nearly, to be humble) everyone.
So now we've found ourselves in the slightly absurd situation of literally inventing how Software 3.0 will work, and so we now need a few sharp builders to help us get there even faster.
What we actually do
Software 3.0 goes way beyond vibe-coding. Agents, subagents, skills, context engineering, evals, guardrails, MCP servers, deterministic pipelines wrapped around wildly nondeterministic models.
Getting an agent to do the right thing once is easy. Getting it to do the right thing every single time, unsupervised, overnight, without burning through API calls like a crypto bro in 2021? That's harness engineering.
And here's the part that keeps us up at night: making agent behavior predictable is the same discipline that's going to train real-life robots. The stuff you'll learn here won't expire when the next model drops, and the people early to this will benefit from a head start that compounds for life.
What you'd actually do
You'd work with our engineering director, our business development director, our operations director, your fellow interns, and silly ol' me (our mismanaging partner and thought needer).
- Build agent harnesses, break them, then build them again so they stop breaking
- Write skills and evals, wire up tools, design the guardrails that make agents behave
- Watch an agent fail in a way no human has ever seen, then make that failure impossible forever
- Absorb a LOT of frontier knowledge, fast, because we will be teaching while we delegate
- Learn everything you'll need to know about marketing and distribution in the years ahead
- Be first to learn how to build Agent-Native Applications (and be encouraged to monetize them)
The trade is explicit: you get the reps alongside us, we download our business expertise into your head through osmosis and questions. Nobody's fetching the coffee. We have robots for that. (jk... mostly).
What you'll learn
How production agent systems are actually engineered: agent loops, tool-use logic, orchestration, multi-agent coordination, memory architecture, model routing, sandboxing, eval datasets, scorers and judges, verification loops, regression tests, constitutional governance, failure-analysis workflows.
Deeper into the architecture we're developing for Agent-Native Applications: portable Agent Operating Models (AOMs), Policy Decision Points (PDPs), constitutional governance layers, ICM memory protocols, operational contracts, permission boundaries, lifecycle controls, conformance testing, and cross-host portability.
These are the disciplines that turn raw model intelligence into reliable, inspectable, governable software, and many of them are only just emerging as distinct roles at frontier AI companies. By the end you won't just understand Agent-Native Applications. You'll have helped build the platform they run on, and have the chance to launch one (or several) of your own.
You, probably
- You're a builder, tinkerer or fixer, and/or a marketer, communicator or persuader. Degree optional, portfolio optional, curiosity mandatory.
- You're smart AND fun to be around. These are not opposites. You're the person in your friend group who's "good at computers" but you also get invited to things.
- You're self-aware. You can say "I don't know" without crying or dying.
- You're ambitious but you don't know where to point it yet. You can feel that something enormous is happening right now and you'd rather be inside it than watching recap videos.
- You like dogs. You gotta like dogs or it just won't work.
Not you
- You need a paycheck asap. Completely fair, genuinely. This just isn't that. (Eventually though).
- You want a task list and zero ambiguity. We're inventing this as we go. Ambiguity is the job.
The deal, plainly
This is an unpaid, educational internship, and we know exactly how that sounds. Here's the difference: we're not scavenging the job market for free labor to save a dying business model. We're experienced entrepreneurs who have built, scaled and sold companies over the years. This one will be the biggest.
So why hire interns if we can afford to recruit almost anyone we want? Fair question, simple answer: unlearning is more expensive. After interviewing several senior engineers, we realized this new Software 3.0 paradigm is much easier to teach to early-career people simply because they don't have as many bad habits that need to be unlearned.
If you prove your capabilities, your problem-solving and your character, we'll be hiring from this group first when paid roles open, and we intend on recruiting a lot of people over the next few years. No promises, just intentions.
Worst case, you spend a few months at the edge of the frontier and walk away with a skillset almost no one has yet. Best case, this one decision completely changes the trajectory of your life and career.
How to apply
Sheesh, no resume drama. We won't even look at it, promise. All we want to know is what you want the next few years of your life to look like.
Submit a video answering three questions. Video is preferred, but if that's not your format, email your answers to team@nickapollo.com instead. We care about the answers, not the production value, and this is one thing we will never outsource to our agents: a human reads or watches every single application.
- 1.The coolest thing you've ever built, figured out, or gotten away with (school projects count, chaos with friends, family and/or mortal enemies counts too)
- 2.What you think an "agent harness" is, in your own words (wrong answers accepted if they're interesting)
- 3.Why you want to join our team
Applications close soon and the internship starts as soon as we find the right people. We move quick.
Before day one you'll sign a confidentiality agreement, an assignment of intellectual property, and an internship agreement. You'll see every document in advance, and if your school gives credit for this we'll do the paperwork.
Recruiting is carried out for more than one independent company and program; the one that would engage you depends on the role and where you fit best, and it's named in the agreement you sign. All applications are subject to our terms of use and privacy policy.
If you smirked at least once reading this and you like dogs, you very well just might be our person.