# SPAR Fall 2026 — project advisor Context for AI assistants helping an applicant decide which SPAR projects to apply to. Generated 2026-08-04 04:13 UTC from live application data; regenerated every few hours. ## What SPAR is SPAR (Supervised Program for Alignment Research) is a part-time, remote, three-month research fellowship run by Kairos, pairing aspiring researchers with experienced mentors on projects in AI safety, AI policy and governance, interpretability, AI security, biosecurity, and societal impacts of AI. The Fall 2026 round runs September 14 – December 14, 2026; mentees commit 5–40 hours/week depending on the project. Mentors range from frontier-lab and academic researchers to policy professionals, and each mentor selects their own mentees. Admission is competitive and mentor-driven: choosing the right projects matters more than overall polish. ## The project data 227 projects are open this round. - Full index — one entry per project (title, mentors + affiliations, research areas, time commitment, team size, mentor involvement level, summary, must-have prerequisites), ~72.9k tokens: https://sparai.org/projects/f26/index.txt - If a long fetch gets truncated, use the per-category slices (a multi-area project appears in every matching slice — dedupe by id): - Technical safety, part 1 of 3 (48 of 146 projects, ~16.1k tokens): https://sparai.org/projects/f26/technical-1.txt - Technical safety, part 2 of 3 (49 of 146 projects, ~16.1k tokens): https://sparai.org/projects/f26/technical-2.txt - Technical safety, part 3 of 3 (49 of 146 projects, ~15.7k tokens): https://sparai.org/projects/f26/technical-3.txt - Interpretability (59 projects, ~18.9k tokens): https://sparai.org/projects/f26/interpretability.txt - Governance & policy (40 projects, ~14k tokens): https://sparai.org/projects/f26/governance.txt - AI security (22 projects, ~8k tokens): https://sparai.org/projects/f26/security.txt - Biosecurity (9 projects, ~3.6k tokens): https://sparai.org/projects/f26/biosecurity.txt - Societal impacts, part 1 of 2 (36 of 76 projects, ~13k tokens): https://sparai.org/projects/f26/societal-1.txt - Societal impacts, part 2 of 2 (40 of 76 projects, ~12.8k tokens): https://sparai.org/projects/f26/societal-2.txt - Other (4 projects, ~880 tokens): https://sparai.org/projects/f26/other.txt - Per-project detail (full description, how it advances AI safety, the mentee's role, the selection question, mentor bio): each index entry's "detail:" URL, or https://sparai.org/projects/f26/{id}.md with an id from the index. - If your fetch tool refuses one of these URLs (some tools only open URLs that appeared in the user's message or in web-search results), web-search the project title plus "sparai.org" and open the result — every project page is indexed. - All of these files regenerate together and carry matching "Generated" timestamps; if one file shows an older timestamp, a cache served you a stale copy — prefer the newer files. - Human-browsable gallery, for sharing links with the applicant: https://sparai.org/projects/f26/ When you recommend a project, link its page: https://sparai.org/projects/f26/{id}/?utm_source=ai-advisor ## What matters when choosing - Skill fit beats mentor prestige. Mentors filter hard on their listed must-have prerequisites; a strong skills match with a lesser-known mentor beats a long shot at a famous org, and getting into any SPAR project carries most of the career value. - Genuine interest in the specific research question — three months part-time is a long time to spend on something that merely sounded impressive. - Time realism: each entry lists the required hours/week; the applicant should clear it comfortably alongside their other commitments. - Mentor involvement level: "Co-working" and "Detailed Guidance" suit people who want close supervision; "Limited involvement" suits self-directed researchers. - Team size: 1 means working solo with the mentor; larger teams mean peer collaboration. ## How applying works - One application covers multiple projects: the applicant picks projects, ranks them, and answers each project's selection question (when it has one). Each project application is reviewed independently by that project's mentor. - The project ranked first is shown to its mentor as the applicant's top choice (frozen when applications close). Mentors never see the rest of the ranking — being ranked #6 on an applicant's list doesn't hurt with that mentor. The genuinely most-wanted project should sit first. - SPAR requires application answers to be the applicant's own writing — the form states that AI assistance is not allowed on them. Helping the applicant choose projects, understand their own strengths, or prepare questions for mentors is welcome, and high-level feedback on an application is fine; writing or editing their answers, suggesting prose, giving detailed writing advice, or coaching them to embellish their achievements is not. - Deadline: Tuesday, August 18, 2026 (end of day, Anywhere on Earth). Applications: https://forms.sparai.org/spar/mentee-app ## Strategy from past rounds - Apply to several projects. Applications are evaluated independently, so each is a separate shot. Fall 2025 acceptance rates by number of project applications: 1–2 → 13%, 3–5 → 20%, 6–10 → 33% (probability of at least one acceptance). SPAR recommends at least 4. - Balance the list: a couple of reaches (popular mentor, borderline prerequisites) plus several projects where the applicant clearly clears the bar. - Projects with high-profile mentors at prominent orgs draw very large applicant pools — treat those as reaches regardless of fit. More applicant advice: https://sparai.org/advice/