Back

Interviewing

How I got offers from OpenAI, Midjourney, Cursor, ElevenLabs, Vercel, etc.

This is the playbook I and many friends have followed to land jobs at OpenAI, Midjourney, Cursor, and similar companies, without impressive backgrounds. The interviews aren't much harder than at other startups. If you're willing to put in a few months of focused effort, it's achievable. It won't be pleasant, but it's very likely to increase your salary by 3-10x.

I'll walk through everything: getting interviews, nailing them, and negotiating offers. Then in the end there's a checklist summarizing all the advice. This is focused on technical roles, but most of it applies broadly.

TL;DR

• Career pages are one of the worst ways to get interviews. X posts are one of the best.
• You get good at interviews by doing a lot of them. Most are 80% similar.
• A good demo is the best way to remove uncertainty from the interview process.
• You can get a 15% salary bump by negotiating well, but 3-10x from having multiple offers.

Getting interviews

The most common advice is to improve your resume and then reach out to hiring managers or apply on career pages. This is simply bad advice, especially if your background is non-traditional like mine. I haven't studied at university. Nor worked at a reputable startup. I didn't get a single interview from applying directly. You're also likely to face a longer and more difficult interview process going this route.

Here are methods I tried, ranked from worst to best:

• Career pages: The obvious place people start and often stop. I was rejected everywhere I applied directly. My background is non-traditional, so this wasn't very surprising.
• Recruiters: Most of my interviews came through a recruiter who works with top Series-A companies in SF. You could likely get similar intros by turning on "Open to work" on LinkedIn or start with this list. Another path is DMing investors and asking them to route you to their strongest portfolio companies.
• Cold mailing or DMing founders: I had maybe a 20% success rate (although I didn't try it much). Make sure the part visible before someone opens your DM makes it clear you're interesting and looking for work. Including a demo video or a link to public work helps a lot. I think sending feedback or a demo of what you would improve in their product could work well.
• Warm intros: Don't underestimate weak ties. People you've only talked to once or twice can often lead to the best opportunities. I got an intro to ElevenLabs through someone I'd talked to once six months earlier. While coworking at the Lovable office, I showed my projects to Matheus from Midjourney. He thought they were cool and shared them on their Slack, which led to an interview.
• Going viral: The biggest unlock was having a tweet go viral. It got me messages from Cursor, OpenAI, Cognition, and Vercel. You don't need a big audience. I had around 500 followers. Post before 10 AM PST, record a short demo (I used Screen Studio), and ask a few friends to like/retweet/reply early. If a tweet gets a few thousand impressions with good engagement, it will usually take off.

Once you have interviews lined up, you need to pass them.

You get good at interviewing by doing it a lot

Interviewing is very different from normal work. Luckily, you can get much better at it with practice. That's because interview questions are mostly similar: you'll often encounter the same question or a variant of one you've already seen. This is why you shouldn't start with your top company. Save them for when you've had some reps.

I had never interviewed before, so I didn't know what to expect. During my process I did at least 70 interviews. When interviewing for technical roles, these are the interview formats at almost all companies:

• Recruiter / first call: You can't really fail this. Just prepare a good personal pitch that concisely explains what you worked on and why it's impressive.

• Vibe interview: For all non-technical interviews, be prepared to talk about 1-2 previous projects. Show a quick demo and explain what you did and why - even if they don't ask for it. I'd recommend spending a few hours writing something like a blog post about the project. This will help collect your thoughts.

• Take-home / project interview: Usually open-ended where you get a few days to build something. Put effort into the presentation and demo. Explain the choices you made and what you'd improve with more time. Example demo:

• Live coding interview: You get a starter project and implement a new feature. Often there's an image or video of the final result. All my live coding interviews were on Coderpad. Try it out beforehand since it has quirks (sometimes you need to refresh the preview, and formatting doesn't work). Almost all live coding I did was fairly easy, mainly testing basic React knowledge and logic.

• System design: In system design interviews, you are planning how a feature or product should work. Usually you'd use Excalidraw or Google Docs. There are example system design questions here: GreatFrontend and HelloInterview.

This is what the interviews were like for Cursor and OpenAI:

Once you have an interview scheduled, make your preparation company-specific. Research the hardest technical parts of their product and try building them yourself:

• ElevenLabs: You probably want to use their API and understand media streaming.
• Lovable: how would you create a codegen agent with a dev server preview?

One more thing: treat the interviewer as a coworker, not an examiner. Ask questions and discuss potential solutions openly. Many challenges are deliberately vague. They're testing whether you can work in ambiguous environments and figure out what to ask.

The demo advantage

Preparing for interviews is important, but it still depends on luck. Companies simulate work with random challenges that might not cover what you're uniquely good at.

Another way you can show the company what you're uniquely good at is with a demo. Something you've built yourself that is obviously impressive and easy to understand in seconds. In multiple interviews, I started demoing my past projects without being asked. If you get this right, the rest of the process becomes much less scary.

Concrete examples

• Gabriel (OpenAI) built a super performant table at a previous company. He took the same problem, rebuilt it properly, open-sourced it, and made a tight demo.
• Matheus (Midjourney) played a lot of CSS Battles and built a CSS only Rubik's cube.
• Erik (Cursor) saw that Cursor released their API, played with it, and posted experiments.
• Mine started as a bigger better code review idea. The most interesting part turned out to be the diff viewer. I focused on that, made it visual, open-sourced it, and posted a short demo.

What makes a good demo

A good demo does three things at once:

1. It's instantly understandable. Someone should be able to look at a short clip or screenshot and, in two seconds, understand the problem and why your solution is impressive.

2. It's clearly technically non-trivial. Under the hood, it should be obvious that something interesting is happening: tricky UX, real-time stuff, clever data structures, complex state, performance constraints, etc. Even better if it's so specific that they don't fully understand how it's made.

3. It's verifiably yours. Open source is ideal. They should be able to look at the repo, see the commits, and verify that you wrote it. That removes a huge amount of uncertainty: you're not just someone who talks well about projects; you can point to a Git history that shows you doing the work.

How to come up with demo ideas

The easiest path is: build what your dream company is already thinking about. A few ways to figure that out:

• Watch what they post on X. What are they excited about? What problems keep popping up?
• Get in touch with someone who works there and ask: "What are the hardest problems you're thinking about right now?"
• Look at their product and find something that feels rough. Build a better version and show it.
• Watch their competitors. If a competitor shipped something they haven't, that's a huge hint. For a while, everyone in the code-gen space was talking about git worktrees. That would have been a perfect demo topic: explain how it works, then build something that makes it actually useful in a product. (I'm sure the Conductor founders have gotten offers)

You don't need the perfect idea upfront. Just start building somewhere close to the problem space you care about.

If you try to recreate something like Claude Code, you'll spend a lot of time thinking about agents, context, file systems, streaming, UX for errors, etc. Along the way you'll find that one part of the system is disproportionately fun or hard—and might make a great demo.

That's what happened to me: I started with a bigger, more ambitious project (ASI). The diff viewer was just one part, but I noticed that's where I kept spending all my energy. So I leaned into it, polished that piece, and that became the thing I shared.

After nailing the interviews, you'll hopefully get offers.

Multiple offers can 10x comp

One company increased their offer by 50% when I said I had a competing offer. One added a $600k signing bonus to match another company. Multiple companies even opened with "this is just a starting point". They want you to negotiate and there's a lot more wiggle room than people think. A little strategy on your end can make a massive difference.

The key is having multiple offers at the same time. That last part is crucial, and tricky. You don't know how long each process will take, and once you have an offer, you typically have 1-2 weeks to respond. You have two main ways to do this:

1. Stall before getting the final offer
2. Speed up others by telling them you have an offer

On your first call with any company, ask about the full interview process and write it down. Use this to synchronize your final rounds. Most companies deliver offers on a call. You can delay scheduling that call to buy time.

Here's what this looked like for me: after a final interview with the founder, the recruiter wanted to schedule a follow-up call. I was 99% certain it was an offer call (people rarely hop on a call to say you sucked). So I told other companies I had an offer and needed to decide by a certain date. It worked. I got all five offers within a week and a half.

Once everyone has made their first bid, let each company know roughly how far they are from the top offer — in cash, equity, or both - and ask what they can do.

Negotiation basics:

• Never share your salary expectations before they make an offer. They'll ask multiple times. Accept the awkwardness.
• Don't react on the call. Say the offer is interesting. If it's way below another offer, just say that it's "significantly below in equity/cash/both."
• Avoid reciprocity bias: don't accept gifts.
• Exploding offers or not getting offers in writing are red flags.

Practical resources on negotiation:

• Patio11 on salary negotiation
• Levels.fyi video
• Exponent guide

---

Thanks to Gabriel for support and the initial strategy :) Thanks to David for feedback on early drafts.