We're now housed at Convergent Research, funded to apply Field Strategy beyond formal methods to catastrophic AI risk more broadly, and growing from one person to twenty. The rename is the finishing touch.
Why “Atlas Computing” originally?
Atlas Computing began work in late 2023, mapping paths (hence Atlas) to technologies that would let humanity work more safely with untrustworthy but highly capable AI, primarily through formal methods (hence Computing). With early support from davidad, the original program director of ARIA’s Safeguarded AI program, we set out to discover how formal methods could best leverage near-term foreseeable AI capabilities.
Our early work focused on both sense-making and field-building, as most experts in formal methods were naturally and understandably skeptical about the utility of transformer-based systems. We designed tools one would want from highly capable AI systems as well as novel, user-friendly approaches to using AI systems to generate provably correct software. We brought together experts from the field and put them in a room with researchers from frontier AI labs at FMxAI, which has grown into an event series (next in London on March 8-10, 2027). We also proposed novel ideas like Formal Scalable Oversight and helped to seed the creation of Lean’s CSLib.
Along the way we helped create three AI security organizations. We scoped three orphaned problems (two in formal methods and one in hardware security), identified compelling solutions that stood up to expert scrutiny, and after recruiting a first-time founder for each, helped them raise their initial capital. This led to Theorem Labs, Oath Technologies, and Earendil.
We’ve since documented our process as Field Strategy, presented it in talks and long-form blogposts, and hope to help effectively route future philanthropic capital to address civilization’s maintenance backlog.
Why “Atlas Ignota” now?
Ignota is Latin for unknown things. We’re now looking at risks and interventions far beyond computing, and focusing on orphaned problems means we’re venturing well beyond what’s known.
Hence: Atlas Ignota.
We still plan to support the formal methods ecosystem, but are now funded to apply our Field Strategy approach to problems ranging from AI alignment and verification for AI pacing agreements to epistemic security and cybersecurity of critical infrastructure. We’re currently hiring for two roles, scaling the team from 1 to 22 as fast as we can hire well (currently at 4):
Founders in Residence pick a single catastrophic AI risk and we provide a home for them to understand the problem, de-risk the solution direction, and elicit conditional commitment from potential users/stakeholders before we introduce them to funders to spin out and lead the resulting effort. We’re specifically sourcing founders to focus on these problems in AI infrastructure, the substrate for alignment, control, security, workload verification, and monitoring). [apply here]
Field Strategists scope multiple orphaned catastrophic AI risks, work out what interventions could mitigate them, and source teams to execute on the strongest interventions. We’re looking for Field Strategists for all of the above topics. [apply here]
If figuring out what important thing is currently falling through the cracks—or building the institution that should exist to address it—sounds like your kind of work, we’d love to hear from you.
We’re grateful to Coefficient Giving, Effective Institutions Project, and the OpenAI Foundation for supporting this next phase of our work.
How to find us
Our new home is atlasignota.org. All our emails will redirect and various accounts will update. You can drop us a line at hello@atlasignota.org.
And if you’ve been following Atlas’s work: thanks for sticking with us. We’re excited to see what we find in the unknown.
Brief clarifying notes on logistics you probably don’t care about
For those interested in exactly what happened organizationally:
Our previous work at Atlas Computing was fiscally sponsored by Metagov Inc. With the refocus, that team wound down and I’m the only one transferring over. Its assets transferred to Atlas Ignota, a new wholly-owned subsidiary of the 501(c)(3) Convergent Research. We had previously established an independent 501(c)(3), but ultimately never used it, and have transferred control of it to Sparrow Infrastructure.