Science Campaigns¶
Trinity isn't just a scheduler front-end — it's been run end-to-end on real scientific and engineering work across ALCF, NERSC, and OLCF systems. This page is a record of that work: the request as asked, what Trinity actually did (including the dead ends and self-corrections), and what came out the other end.
Every entry below is backed by a full execution trace and, for auditability, a provenance record of why each decision was made — the same "not just the agent said so" standard described in What Trinity is. Where an original prompt wasn't captured verbatim, the page says so plainly rather than presenting a paraphrase as a quote — and partial, mixed, or negative results are reported honestly, because a campaign that catches its own mistake is a better proof of rigor than one that just reports success.
-
Trinity runs
Campaigns run live through Trinity's chat interface — DFT, molecular dynamics, ML benchmarking, and infrastructure verification, each with a real execution trace.
-
Autonomous science campaigns
A systematic study of 14 physics/chemistry/engineering campaigns across DOE systems, probing how well an autonomous agent self-corrects and knows when to trust its own result.
-
Engineering case studies
Porting and performance work — moving code across GPU vendors, fixing GPU kernels, building software from source — with the debugging story and validated numbers, not just a final checkmark.
-
Paper reproductions
Rebuilding the software, rerunning the calculation, and checking the numbers against what was published — one of the best stress tests of both a method and a workflow.
Have a campaign to add?¶
This page is curated from Trinity's own execution traces and campaign reports. If you've run something interesting through Trinity and want it featured here, open an issue on trinity_hub.