Advice for applying

Apply to projects that match your skills.

Mentors often seek very specific expertise—data science, economics, a particular programming language, technical writing, or fine-tuning experience. Brilliant applicants may not land a role if their skill set doesn’t match the project, so focus on opportunities where you already bring strong experience. Use project keywords such as “PyTorch” or “law” to identify mentors explicitly asking for those skills.

Apply to less competitive projects.

In Fall 2025 our most popular project received 249 applications, while the least-applied-to project had five. The Fewer applications badge highlights projects that are less competitive right now, and you can filter by that tag—though those spots can fill up fast once the badge draws more interest.

High-profile mentors and prestigious organizations usually attract larger applicant pools, so if you’re excited about those projects, apply, but also consider other options so you’re not relying on a single outcome. Any SPAR project offers valuable experience—matching your skills and goals matters more than the mentor’s profile.

Apply to lots of projects.

Every project application is evaluated independently, giving you additional chances with each submission. Our Fall 2025 acceptance rates by application count were:

• 1–2 project applications: 13%
• 3–5 project applications: 20%
• 6–10 project applications: 33%

Those figures describe acceptance into any project—applicants who sent 6–10 applications had a 33% chance of matching with at least one mentor. We recommend against pinning your hopes on a single project and encourage you to apply as much as your time and energy allow.

Application questions.

For the general questions, be specific. Instead of stating “ML experience,” describe the tools, models, and projects you’ve worked with. Rather than “I follow discussions and read papers,” cite concrete papers or topics that guide your interests so mentors understand your foundation.

For project-specific prompts, answer every part carefully and back your responses with evidence. Strong answers explain why your solution works in addition to what it is. If you describe an experiment, clarify the rationale, anticipated data, interpretability, and how the insights would advance AI safety.

Strengthening your resume.

There’s no one-size-fits-all recipe; mentors evaluate applicants using their own criteria. Technical projects typically look for strong coding, ML, and model intuition, while policy projects prize well-honed writing. Previous research experience is always advantageous, but SPAR also welcomes diverse backgrounds like biology, philosophy, or multilingual fluency.

Technical applicants can boost their work by completing the ARENA curriculum and doing independent research (paper replications are a solid start). Policy applicants should gather strong writing samples—consider choosing an AI policy topic for a class paper or starting a blog. Holden Karnofsky’s Learning By Writing method is worth exploring.

On rejection.

Rejection isn’t a judgment on your potential to contribute to AI safety or your chances in future SPAR rounds.

• We can’t take everyone who would be an excellent fit. Mentors often have more strong applicants than open slots, so the difference between accepted and rejected candidates can be small.
• Mentors search for specific skill sets. A brilliant applicant might be declined because their strengths lie somewhere else.
• Each mentor decides differently. The mentor, not the SPAR team, chooses whom to shortlist, and different rounds feature different people, so you’ll get a fresh opportunity next time.

In some cases, rejection signals a need to upskill before reapplying, but many applicants do a lot of things right and simply need another shot at finding their fit. We wish you the best luck in applying. Feel free to reach out to [email protected] if you have any questions