About Sequent
Sequent is a new AI alignment research organization, founded in 2026 by researchers from the UK AISI’s Alignment Team and Timaeus. We aim at higher a priori confidence in aligned outcomes by pursuing a portfolio of theory and empirics bets, any one of which, if it succeeds, would meaningfully advance the field. We invest heavily in research automation to accelerate progress, and we believe that theory unlocks higher automation: more principled approaches give us better filters for which directions of automated research are promising.
For more information, see our announcement.
About Our Research
We work across a portfolio of research areas, currently including:
- Scalable oversight: empirical work on protocols (debate, recursive reward modeling, prover-verifier games) that allow weaker overseers to supervise stronger systems, paired with complexity-theoretic work on equilibria and reachability.
- Complexity theory: theoretical work on modeling the interaction of superintelligent agents with lower complexity training environments, including applications to scalable oversight, heuristic arguments, and other agendas.
- Learning theory: singular learning theory and its applications, deep learning theory, computational mechanics, etc.
- Personas: theory and empirics of low-dimensional structure within model behavior across training and token dimensions.
The full set of bets has not yet been finalized but, in the future, may include further agendas like:
- Heuristic arguments: mechanistic understanding of what models know, low-probability estimation.
- Game theory: mechanism design, agent foundations, open-source game theory.
A cross-cutting focus is Research Automation: building infrastructure and tooling to scale all of the above by leveraging AI research assistants at every level of the stack, across both theoretical and empirical work.
About the Team
We’ll soon be hiring Research Scientists across our research programs. The team you’d join depends on your strengths and interests. In your application, indicate which program(s) you’d be most interested in.
About the Role
As a Research Scientist at Sequent, you’ll work as part of a research program led by a senior researcher (which could be you!). At Sequent, our goal of reaching higher confidence in alignment will require significant new contributions in both methods and scientific understanding. The work you do day-to-day will be part of a collaborative research program; all programs involve deep technical research, with strong support for cross-program collaboration. We welcome both empirical and theoretical profiles.
Responsibilities
- Research within your program: both translating and modelling alignment problems into concrete empirical or theoretical form and executing on the resulting concrete problems. Execution means designing and running experimental protocols and methods for empirics, and proving, conjecturing, and investing in autoformalization for theory.
- Writing and presentation of completed research in the form of papers, blog posts, and talks.
- Communication of research progress and obstacles to members of your team through channels like Slack on a daily basis and in weekly meetings.
You May Be a Good Fit If You
- Have a graduate degree (Ph.D.) or equivalent experience in a field relevant to the program you’re applying to: ML, CS, mathematics, physics, statistics, philosophy, or something closely related
- Have a track record of research and strong technical writing ability: papers, preprints, or comparable output
- Have a strong mathematical background, even if your work is primarily empirical
- Can credibly articulate why your area of alignment work is worth pursuing
- Are willing to use AI tools aggressively in your own workflow, with appropriate care to not get fooled!
- Are motivated by alignment of artificial superintelligence (ASI) and want to contribute to it full-time
Strong Candidates May Also Have
- (For scalable oversight) Hands-on experience with debate, prover-verifier games, RLHF empirics at frontier scale; familiarity with the scalable oversight literature; background in game theory, complexity theory, mechanism design, or decision theory
- (For complexity theory) Demonstrated expertise in theoretical computer science, including creative modeling and assumption generation
- (For learning theory) Background in algebraic geometry, Bayesian statistics, information theory, statistical physics, optimization theory, or learning theory; familiarity with (singular) learning theory
- (Across programs) Experience scaling experiments to billion-plus parameter models; a track record of productive collaboration with engineers; prior engagement with our research or sibling organizations (Simplex, ARC, etc.). Experience with AI alignment is a plus, but is not a requirement: we are excited to bring experienced researchers from other fields into alignment work, and we see our ability to facilitate that transition as one of Sequent’s comparative advantages.
Research Program Lead
We’re also building out program leadership across our research areas. We expect to fill these roles primarily through targeted headhunting rather than open applications. If you think you might be a strong fit for a Research Program Lead position, a senior researcher who could direct an entire research program at Sequent, please reach out to us directly.
Logistics
- Location: Berkeley strongly encouraged; London is a secondary hub; remote may be considered in exceptional cases.
- Visa sponsorship: Yes, for relocation to Berkeley
- Start date: Rolling
Expression of Interest
We expect to open a full hiring round soon. In the meantime, if you’re interested in roles at Sequent, please fill in this Expression of Interest form.