TL;DR: AE Studio is running a new government-funded research program that pairs alignment researchers with our internal engineering team for two-week sprints. We will select six alignment researchers whose proposed work attempts to answer one of our six open questions. Promising sprints have the opportunity to expand to longer funded work.
Researchers are asked to commit two hours a week, and in return AE will provide compute and 60+ hours of engineering from us. Apply here.
Applications close 11/15 at 11:59 PM PST.
What is CRUCIBLE?
CRUCIBLE is an IARPA-funded research program executed by AE Studio. The premise is that alignment researchers have more ideas than they have the time or resources test. AE Studio has been operating a model where we match an external researcher with an internal engineering and ML team, so that the research can spend as little as two hours a week and still get high-quality results.
CRUCIBLE will focus on six research questions, drawn from priorities set in federal legislation for fiscal year 2027. We are soliciting research proposals against those questions.
We will select six researchers and start sprints on a rolling basis, with the first beginning on November, 23rd.
The research questions
Pick the one your research idea best:
Do models internally represent alternative hypotheses that they fail to report? Is the competing explanation present in activations and suppressed at output, can it be surfaced? Does sycophancy to the prompt’s framing drive the suppression?
What signals, internal or behavioral, predict that an AI-assisted judgment is wrong in an escalatory direction? Across analytic, classification, and operational uses, which model-side signals (uncertainty, inconsistency across framings, suppressed alternatives, attribution to the wrong evidence) correlate with errors that would raise rather than lower escalation risk, and which of them can be read in time-constrained settings?
What can go wrong in a multi-agent AI system cannot go wrong with a single agent? When two or more AI systems communicate, delegate, negotiate, or compete, what failure modes appear that have no single-system analogue? We want to know what is new, not just what is amplified.
Can AI systems coordinate in ways their operators cannot see? Under what conditions do interacting systems develop shared strategies, hidden signals, or joint deception, and what would it take to detect that from outside?
Do an agent's motivations produce any behavioral signature at all? i.e Can two agents with different objectives but the same capabilities and persona be told apart from their behavior in interaction? If so, is that signature stable across model families and across changes to the agent’s prompt, persona, or memory, or does each configuration look different from the outside?
What makes an agent's loyalty to its user hold? Which agents resist being turned, and is the difference in training, in reasoning, or in what the agent knows about who it’s talking to?
If you have an idea that sits between or just outside of these questions you are still welcome to apply and make the case. We will read every application.
Who should apply?
Researchers with a strong viewpoint on one of the research questions, but don't have the bandwidth or resources to run the experiments solo.
University, company, lab, or independent applications are all welcome. Early-career and senior researchers alike; good ideas come from anywhere.
How it works
Researchers commit to:
~2 hours/week for the duration of the 2 week sprint
A research hypothesis pointed at one of our questions
Participation in the initial planning session, updates throughout the sprint, and demo sessions at the end of each week
Your expertise in interpreting results
AE Studio commits to:
A fully managed AI engineering team running the sprint
Cloud compute from AWS, GCP, Azure, and specialized ML platforms
API access to frontier models for evaluations, synthetic data generation, and related tasks
Working code and documented results
A structured analysis and final report
After the sprint:
Two weeks may not be enough to know an idea is worth a year, but it is should be enough to know whether it deserves another few months. Promising sprints will have an opportunity to extend an additional 1-6 months depending on research needs and your availability. There is no obligation to extend.
The application
On the application you will pick a research question and write a 500-word abstract addressing three questions:
What are you trying to do?
What is the potential impact if fully validated?
Could the idea feasibly be tested within two weeks?
We will score on novelty, feasibility, and impact against the chosen question, then the strongest applicants will be invited to a short call to talk through the idea. Because we are starting sprints quickly we will make some selections on a rolling basis before the deadline. Applying early helps.
All applicants will be notified by 12/4. Selected researchers should be able to start within two to three weeks of notification.
FAQ
How much time commitment is this? About two hours per week per sprint: a planning session, async updates, and a demo at the end of each week. If the work extends, the cadence stays similar.
Do I have to address one of the six questions? Yes, that is the point of this solicitation. If your idea is adjacent, apply and make the case.
Is there compensation? Yes. Researchers receive a $1000 stipend per two-week sprint. Extended engagements are compensated separately.
Who is funding this? CRUCIBLE is funded by IARPA as fundamental research.
Can I participate if I'm at a university or company? Yes, if you can enter a subcontractor agreement with AE Studio.
What compute is available? Cloud compute from AWS, GCP, Azure, and specialized ML platforms, plus API access to frontier models.
What happens after the first sprint? We produce a final report with documented results. If there is signal, we recommend an extended engagement of 1-6 months.
What is the selection process? We review as applications come in, invite promising applicants to a short call, and notify everyone by 12/4.
Apply
Apply here. Applications close 11/15 at 11:59 PM PST.
If you know someone who should see this, please send it along. The talent pool for these questions is small and we want to reach all of it.
CRUCIBLE is funded by IARPA and executed by AE Studio. The views, opinions, and findings contained herein are those of the authors and should not be construed as representing official policies or endorsements of IARPA or the U.S. Government.
TL;DR: AE Studio is running a new government-funded research program that pairs alignment researchers with our internal engineering team for two-week sprints. We will select six alignment researchers whose proposed work attempts to answer one of our six open questions. Promising sprints have the opportunity to expand to longer funded work.
Researchers are asked to commit two hours a week, and in return AE will provide compute and 60+ hours of engineering from us. Apply here.
Applications close 11/15 at 11:59 PM PST.
What is CRUCIBLE?
CRUCIBLE is an IARPA-funded research program executed by AE Studio. The premise is that alignment researchers have more ideas than they have the time or resources test. AE Studio has been operating a model where we match an external researcher with an internal engineering and ML team, so that the research can spend as little as two hours a week and still get high-quality results.
CRUCIBLE will focus on six research questions, drawn from priorities set in federal legislation for fiscal year 2027. We are soliciting research proposals against those questions.
We will select six researchers and start sprints on a rolling basis, with the first beginning on November, 23rd.
The research questions
Pick the one your research idea best:
If you have an idea that sits between or just outside of these questions you are still welcome to apply and make the case. We will read every application.
Who should apply?
Researchers with a strong viewpoint on one of the research questions, but don't have the bandwidth or resources to run the experiments solo.
University, company, lab, or independent applications are all welcome. Early-career and senior researchers alike; good ideas come from anywhere.
How it works
Researchers commit to:
AE Studio commits to:
After the sprint:
Two weeks may not be enough to know an idea is worth a year, but it is should be enough to know whether it deserves another few months. Promising sprints will have an opportunity to extend an additional 1-6 months depending on research needs and your availability. There is no obligation to extend.
The application
On the application you will pick a research question and write a 500-word abstract addressing three questions:
We will score on novelty, feasibility, and impact against the chosen question, then the strongest applicants will be invited to a short call to talk through the idea. Because we are starting sprints quickly we will make some selections on a rolling basis before the deadline. Applying early helps.
All applicants will be notified by 12/4. Selected researchers should be able to start within two to three weeks of notification.
FAQ
How much time commitment is this? About two hours per week per sprint: a planning session, async updates, and a demo at the end of each week. If the work extends, the cadence stays similar.
Do I have to address one of the six questions? Yes, that is the point of this solicitation. If your idea is adjacent, apply and make the case.
Is there compensation? Yes. Researchers receive a $1000 stipend per two-week sprint. Extended engagements are compensated separately.
Who is funding this? CRUCIBLE is funded by IARPA as fundamental research.
Can I participate if I'm at a university or company? Yes, if you can enter a subcontractor agreement with AE Studio.
What compute is available? Cloud compute from AWS, GCP, Azure, and specialized ML platforms, plus API access to frontier models.
What happens after the first sprint? We produce a final report with documented results. If there is signal, we recommend an extended engagement of 1-6 months.
What is the selection process? We review as applications come in, invite promising applicants to a short call, and notify everyone by 12/4.
Apply
Apply here. Applications close 11/15 at 11:59 PM PST.
If you know someone who should see this, please send it along. The talent pool for these questions is small and we want to reach all of it.
CRUCIBLE is funded by IARPA and executed by AE Studio. The views, opinions, and findings contained herein are those of the authors and should not be construed as representing official policies or endorsements of IARPA or the U.S. Government.