Watch a recorded Run in the real GUI.

Pick a scenario and open its replay. It is the Traceome GUI itself, reading a finished Run. Drag the replay slider along the bottom (wide screens) and you see only what was on the record at that moment: the stage, the questions the researcher was asked, and each answer with its evidence.

The demo Run

1. 

   The demo Run · #457 · earlier build A human genome, 29 targets, from raw reads to the report

   HG001, a public human reference sample, and a request of 24 rsIDs, BRAF V600E and other targets. Merged IDs are followed, a gene-change name is placed, and the researcher amends two coordinates. 33 Work items: 18 observed, 8 not carried, 7 no VCF record.

   2026-10-29 · build d3f3c340 · nvidia/Nemotron-3.5-Super-VL · HG001 · 457-human-20261029T084406Z-d3f3c340-a1

   A Run of an earlier build, shown as it originally ran, with its original report layout. The link opens at the report.

   R2 R17

The latest Runs

Recorded development Runs, shown as they are, newest first. Each card names its build; only the first is on the final build.

1. 

   Final live Run · paper From a paper citation to a report, on the final build

   The Mullins 2020 APOE paper and an HG002 human sample: the Run finds and reads the paper, takes the reads from Object Storage, measures its health and writes the new report.

   2026-10-30 · build 32b72b74e · nvidia/Nemotron-3.5-Super-VL · HG002 (about 1×) · 4-mfinal-mullins-20261030T141805Z-32b72b74-a1

   Still running when this page was written: Execution places the prebuilt reference index (about an hour), then the H100 job runs. No post-Run review yet. At about 1× depth, genotypes are not expected to be established.

   R12 R19 R20

2. 

   Run S · smoke Run health and coverage, from staged reads

   A C. elegans sample read straight from Object Storage with a prebuilt reference index. The Run measures itself: 97.6% of reads mapped, depth 17.1, and each target's own depth.

   2026-10-30 · build 028a063a7 · nvidia/Nemotron-3.5-Super-VL · SRR7145496 (C. elegans; identity unresolved) · 4-msmoke-celegans-20261030T132733Z-028a063a-a1

   Older report layout. One wording defect on its page ("the paper researcher") was fixed in the final build. Biological correctness of this sample is unresolved R13

   R19 R20
3. No GUI served for this Run
   Run M · paper The first full paper Run

   Mullins 2020 + HG002, before the last fixes: the paper found from a bare citation, three variants read from it, and a report with a "What papers found" section.

   2026-10-30 · build a3cd32a23 · nvidia/Nemotron-3.5-Super-VL · HG002 (about 1×) · 4-mlive-mullins-20261030T095056Z-a3cd32a2-a1

   Older report layout. Each paper target is labelled by its whole paragraph, and table passages name the wrong section.

   R12

Earlier scenarios

1. 

   Run Q · #141 The whole pipeline

   One question about 24 DNA variants, from raw reads to a checked report.

   2026-10-18 · Qwen/Qwen3.5-397B-A17B (comparison baseline) · HG001, a public human reference sample · 141-full-20261018T095210Z-5b06c210-a2

   Run Q used the Qwen comparison model, not Nemotron. Run H used NVIDIA Nemotron on the same reads; its 29 Work items equal Run Q's item for item R5

   R5 R11
2. Screenshot comes with the replay
   Run H · #387 The whole pipeline, on Nemotron

   The same question, answered with NVIDIA Nemotron 3.5 Super VL on Nebius Token Factory.

   2026-10-22 · nvidia/Nemotron-3.5-Super-VL · HG001 · 387-human-a2-20261022T204321Z-7be9123c-a1

   R1 R2 R3 R4 R6 R7 R8 R9 R14
3. Screenshot comes with the replay
   Fork · #437 Skip the GPU: fork a finished Run

   Start again after the GPU step: planning, approval and GPU time skipped, 29/29 answers the same.

   HG001 · 437-fork-8d30b415-a1

   R10
4. Screenshot comes with the replay
   Fork · #369 Correct an answer: amend two targets

   A Fork amends two requests; both old answers change to observed.

   HG001 · 369-fork-0f89a89e-a1

   R10 R17
5. Screenshot comes with the replay
   Run C · #387 A worm genome, and an open question

   Four positions in a C. elegans sample: no VCF record at any of them, and the sample's identity is still in question.

   2026-10-22 · nvidia/Nemotron-3.5-Super-VL · SRR7145496 (C. elegans; identity unresolved) · 387-celegans-a2-20261022T193856Z-59c33dda-a1

   Biological correctness is unresolved: metadata, run alias and BAM disagree on which line this sample is R13

   R13 R7

What you will see in Run Q

Screens from the Run Q replay, at the moments the How it works page describes.

Replay versus live