Before anything else, make sure you are interviewing where you think you are. “General Intuition” is a generic phrase, and three different companies sit near it in search results: this frontier AI lab, the medical-robotics firm Intuitive, and Applied Intuition, the autonomy-simulation tooling company. The one covered here was spun out of Medal B.V., the app that lets gamers record and share clips, around October 2025. Its pitch is that Medal’s enormous library of gameplay video, paired with the controller inputs that produced it, is training data no one else has, and that this teaches a model to reason about movement through space and time.
What General Intuition is, and why it is hiring
The company builds two kinds of model. An action model decides what to do next; a world model predicts what will happen if you do it. Its public flagship is MIRA, a playable multiplayer world model trained on roughly 10,000 hours of Rocket League data from publicly available bots, running at 20 frames per second, built with the research group Kyutai in collaboration with Epic Games. The founding team is Pim de Witte, a Medal co-founder who leads the company, alongside researchers Eloi Alonso, Adam Jelley, and Vincent Micheli, whose background is in world modeling and simulation. The lab is in-person in New York City, with secondary research locations in Europe. It has not disclosed headcount.
The funding is steep and fast. General Intuition launched in October 2025 with a $133.7 million seed round led by Khosla Ventures and General Catalyst, and by late September 2026 it said it had raised over $650 million in a year, most recently a $220 million round at a reported $6.2 billion valuation. The data engine under all of it is Medal’s: TechCrunch reports the app generates about 2 billion videos per year from roughly 10 million monthly active users, each one a record of what a player did and what happened next.
Understand the thesis before you walk in. Most labs building world models plan to sell the model; General Intuition says it wants to keep the agents and sell what they do, starting with AI features for video games and, as its October 2025 launch coverage described, autonomous search-and-rescue drones further out. A candidate who treats this as a text-model company will land wrong in the first ten minutes. The AI-native company interview guides hub collects its nearest peers, and the closest in spirit are embodied-intelligence labs like Physical Intelligence and Skild AI rather than chat-model shops.
The roles open, and what each loop probably probes
General Intuition and Medal share one Ashby careers board, which in November 2026 listed roughly nineteen openings across the two, in-person in New York. Because the roles are in New York City, those listings carry real salary ranges, which is more than most research-lab guides can tell you. The board tags each role by department rather than by company, so sort out which family your req belongs to first; it decides what the interview digs into and, as the bands below show, what it pays.
| Department on the combined General Intuition / Medal Ashby board (NYC, in-person, read Nov 2026) | Representative live titles | Posted base range + equity | What to prep for this family (loop format unconfirmed) |
|---|---|---|---|
| Research / research engineering | Member of Technical Staff – Research; MTS – Simulation; MTS – Infrastructure | $300K-$600K + equity ($300K-$500K for infrastructure) | World models, reinforcement learning, policy learning, simulation, large-scale training, evals for an agent that acts in an environment |
| Applied research / safety | Applied Robotics Engineer; Head of Safety | $180K-$300K + equity | Deploying agents onto real hardware, model and deployment safety |
| Engineering | Senior/Staff SWE – Backend, Frontend, Site Reliability & Infrastructure; Fullstack – Data Platform; Data Security Engineer | $180K-$275K + equity; $180K-$300K for data-platform and data-security roles | Distributed systems, the commercial API, streaming gameplay data at scale, multi-tenant serving, scaling a live product |
| Interns / new grads | Technical Interns and New Grads | $4K-$15K per month | Project-based research or engineering work; study solid data structures and algorithms plus one reinforcement-learning or world-model concept you can explain out loud |
Turn the unknown into questions for the first call
General Intuition publishes no interview loop, and a search across Glassdoor, Blind, and Reddit turned up no verified reviews. The one concrete, prep-relevant datapoint is a single self-reported account on an interview-aggregator site describing a hiring-manager conversation and an AI-assisted coding round it rated hard. Treat that as thin and unverified, but it is the only first-hand signal out there, so prepare for the real possibility that the coding round lets you use a model. Beyond that, do not walk in expecting a known sequence. A research lab at this stage usually runs something like a recruiter screen, a technical screen, a deeper research or systems round, and a final set of onsite conversations, but that is typical for a frontier lab of this size, not a documented General Intuition process. Every role on the board is in-person in New York with no remote option listed, and I could not confirm whether a lab this small sponsors work visas, so if you need sponsorship raise it on the first call. Make the recruiter screen earn its keep and write the answers down:
- How many rounds total, and what is each one (coding, research, systems, founder)?
- For a research role, is there a paper discussion, a take-home, or a live modeling problem?
- For an engineering role, is the coding a live screen or a take-home, how long, and is it AI-assisted (are external tools expected, allowed, or judged)?
- Which round carries the most weight, and what has sunk strong candidates in it?
- Will I meet a founder, and who is the hiring manager?
- Is the whole loop onsite in New York, is any part remote, and does the role sponsor a work visa?
How to prep the research track
Prep the substance below regardless of the format, since no one has documented the loop. For the Member of Technical Staff roles, the weight is on how an agent learns to act. Be ready on reinforcement learning, world models and model-based control, and the specific problem of learning a policy from logged gameplay rather than live interaction. Know how you would build an eval for an agent whose job is to act well in an environment, not to answer a prompt, and be ready to reason about where a world model’s predictions drift and how you would catch it. Come in with an opinion about why gameplay video paired with controller actions is a strong training signal and where it breaks down as you push toward the physical world. If the recruiter confirms a research round or a paper discussion, tune to the subfield of the team you would join; a research lab like Sakana AI interviews on adjacent ground if you want more reps.
How to prep the engineering track
Again, prep the substance regardless of the format. The product and infrastructure roles are a systems job wrapped around a research lab. The infrastructure MTS role is about training at scale: GPU fleets, data pipelines that move billions of video clips, and serving a world model fast enough to be playable at 20 frames per second. The product engineering roles are about the commercial API, the game-recording pipeline that feeds the data engine, and keeping a live service responsive under load. If the recruiter confirms a system-design round, expect it to lean toward data and model serving rather than a generic social feed: how you would stream and label gameplay data at that volume, or serve a latency-sensitive model to many partners. A coding screen is plausible too, and if it turns out to be AI-assisted, practice driving a model toward a working, well-reasoned solution rather than hiding that you leaned on one. Either way, have one concrete story about a pipeline or serving system you scaled and the exact bottleneck you moved.
What it pays, and how to check
This is the rare lab guide where the numbers are not guesses. Because the roles are based in New York City, the Ashby board posts a base range on each listing, and they are wide but real: Member of Technical Staff research and simulation roles show $300,000 to $600,000, the infrastructure MTS role shows $300,000 to $500,000, and senior-to-staff software engineering roles run $180,000 to $275,000, with data-platform and data-security roles reaching $300,000. Every role adds equity, and the intern and new-grad postings list $4,000 to $15,000 per month. Those are base ranges the company posted, not an invented band, but they are still wide, so the recruiter screen is where you pin your level’s actual number and the equity grant that sits on top of it.
The $6.2 billion valuation deserves a caveat in the pay conversation, because it shapes the equity math: the company announced that figure itself in a thread on X, relayed by RuntimeWire, not through a disclosed priced round, and it did not say whether the number is pre- or post-money. So the equity is where the negotiation lives, and the questions are specific. Ask the grant’s value at the most recent preferred price, the vesting schedule and cliff, the exercise window if you leave, and whether any secondary sale has let employees realize value yet. A raw option count means nothing until you know the strike and the last round’s price. Model the whole package with the total comp calculator, and anchor any salary negotiation on your level and comparable offers rather than the valuation headline.
The real variable here is pace and conviction. A company that raised over $650 million in a year, in person in New York, around an approach that differs from its rivals, is not a calm place, and the interview will test whether you believe the bet. Ask the team you would join what a normal week looks like and how decisions get made when the research is moving weekly. For the behavioral rounds, keep your stories tight with the STAR structure, split your prep with the study plan generator between a systems or research story and the track-specific depth above, and check the AI-startup interview difficulty index for where a lab this young tends to set the bar.
Practice the behavioral round:
