Corpus 1,272 assessed · 1,173 scored · 643 reproduced ≥75 · 168 flagged ·∅ 74.1/100
← New search

Evolutionary repair: Changes in multiple functional modules allow meiotic cohesin to support mitosis.

PLoS Biol · 2020
L1 90/100 PQI 94
Why this verdict

The main results reproduced: recomputed values matched the published ones within tolerance.

Reproduced on the brainbox compute brainarbeit.com
Scoring basis — itemised

Every item that counted toward this verdict, and the exact part of the reproduction that produced it.

Supporting (toward a concern)
Content-critical question only partially held
+2 pts
From: Q5 · Derivability / plausibility 🟡
Content-critical question only partially held
+2 pts
From: Q8 · Severity of the miss (overall human judgment) 🟡
Minor / cosmetic deviation
+1 pts
From: Q3 · Location of the main deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q4 · Cause of the deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q6 · Severity of the deviation 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +5
✓ What held up
  • Same input data as the authors
  • Reported values were directly comparable
  • The central claim held under reproduction
What did not (or only partly)
  • 🟡A deviation arose in the data or preprocessing
  • 🟡A deviation was attributed to the published material
  • 🟡Reported values were not (fully) derivable from the shared data
  • 🟡The deviation was non-trivial in magnitude
  • 🟡Overall, the reproduction showed a material discrepancy
How its reproducibility compares
90/100
Reproducibility score
0.9 SD above mean
vs. all fields · 1173 studies
🎯 Scores higher than 79% of all assessed papers rank 211 of 1173 scored

A 0–100 reproducibility-quality score from the per-question grades, shown as a z-score: standard deviations above (+) or below (−) the mean of comparable assessments.

Reproduction agent’s raw note

DESCRIBED WELL ENOUGH and reproduced 1:1. The paper's central pipeline-derived result is a per-population list of fixed (>=90%) non-synonymous adaptive mutations from pooled-population WGS vs ancestor (S-Table, Fig 3B/3C). We reproduced this for 4 of 15 generation-375 populations (P4,P10,P13,P15; SRA E3-P*), chosen to span every named functional module, using the koschwanez/mutantanalysis logic (bwa->samtools->VarScan somatic vs ancestor E0-11) with current builds + snpEff R64-1-1 annotation (legacy Python2/bwa-aln/GATK3 binaries substituted; reference Ensembl R64-1-1 == S288C). RESULT: 21/21 reported SNPs reproduced EXACTLY (identical ref>alt and frequency), incl. cohesin SMC1/SMC3 (missense), separase ESP1 (missense), and Cdk8/mediator SSN2/SRB8/BOI2 recovered as stop_gained -- directly confirming the paper's 'early stop codons in the Cdk8 module'. 2/2 indels reproduced as the correct frameshift/gene (ALT repr differs due to Excel-mangled table + left-alignment). A systematic +1 coordinate offset (paper 0-based vs VCF 1-based) is present; allele identity and frequency match. ONE flagged discrepancy for human review: the recurrent TIR1 chrV:175658 indel reported at ~96% in 6 populations is only ~5-7% in the raw reads (present in ancestor too), so not called somatic -- possibly an artifact of the paper's GATK indel-realignment (which we omitted) in a repetitive subtelomeric gene; flagged, not asserted as fabrication. NOT ATTEMPTED (out of scope / optional 20%): the remaining 11 populations and the gen-1750 set, wet-lab phenotypes, ChIP-seq, backcross-segregant linkage, plasmid-borne REC8 E148Q (plasmid absent from genome ref), and exact rerun of the legacy GATK3/HTML classifier.

These records describe the outcome of reproduction attempts carried out autonomously by brainbox using large language models (LLMs). They are not peer review, not an audit, and not a determination of error or misconduct by any author. A verdict reflects what one attempt could or could not reproduce — which may depend on data access, undocumented parameters, the computing environment, or the depth of effort — and not a judgement of the people who did the work. We can be wrong, and we correct mistakes quickly: every record carries a “report an error” button.

Assessment versions

Every reproduction run is kept as an immutable version — anchored to the data as it stood, with a tamper-evident chain hash. A rerun (e.g. after an author updates a deposit) adds a new version; the previous one stays on record.

  1. v1 current initial assessment Score 90
    assessed: 2026-06-15 ⛓ 7e2ccc98b2f7
✎ I am an author of this paper

Updated or fixed a deposit, or is there an erratum? Ask us to re-run the metrics. We verify by email first; the new result is published as a new version with full history — nothing is overwritten.

Reason for the rerun

We email you a confirmation link first. The rerun is an objective re-measurement — it cannot change the verdict in your favour, only ask us to look again.

Provenance — full disclosure

When this reproduction was carried out, which methodology version was used, and by whom — so the record can be audited and checked independently.

Reproduced
2026-06-15
Rubric version
v1.0
Assessed by
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-15
no human curator yet
Last updated
2026-08-05

Provisional, curator- or AI-assessed, and independently checkable. A reproduction outcome states what one attempt could reproduce — not a judgement of the authors.

Deep full-text extraction

Model: opus
Founding hypothesis

How do cells adapt when a protein is forced to participate in a different biological function? The paper tests this by forcing budding yeast to use the meiosis-specific kleisin Rec8 in place of its mitotic paralog Scc1 during the mitotic cell cycle, then evolving populations to ask what mutations repair the resulting defects.

Core claims
  • Replacing the mitotic kleisin Scc1 with the meiotic kleisin Rec8 impairs sister chromosome cohesion, advances genome replication timing, and reduces reproductive fitness by 45%. finding
  • Cells adapt to use an existing protein in a new biological function mainly by mutating the protein's interacting partners (mediator complex, cohesin-related genes, cell cycle/S-phase regulators) rather than the protein (Rec8) itself. finding
  • Adaptive mutations restore sister chromosome cohesion and delay genome replication in Rec8-expressing cells, returning replication timing toward the wild-type pattern. mechanism
  • Experimental evolution of Rec8-dependent yeast reveals genes that functionally interact with kleisin and a new link between genome replication timing and sister chromosome cohesion. finding
  • Rec8 protein is more unstable than Scc1 in mitosis, and this instability depends on separase activity. finding
  • Engineering mutations that reduce replication origin firing or slow replication forks improves the fitness of Rec8-dependent cells. method
  • Use of conditional P_GAL1-SCC1 with P_SCC1-REC8 to study acute effects of kleisin replacement in a single mitotic cell cycle. method
  • The sister chromosome cohesion defect in Rec8-expressing cells activates the spindle checkpoint (Mad2-dependent), prolonging mitosis. mechanism
Experimental setups
Assay System Perturbation Readout Platform
Competitive growth fitness assay S. cerevisiae P_SCC1-REC8 strain (mCitrine-marked) vs wild-type Scc1 strain Rec8 expressed in place of Scc1 relative fitness from change in strain ratio over generations
Experimental evolution 15 parallel S. cerevisiae populations expressing Rec8 in place of Scc1 Rec8 substitution; serial propagation fitness increase over 1,750 mitotic generations
Whole-genome sequencing evolved S. cerevisiae populations none (post-evolution genotyping) adaptive mutations (mediator complex, cohesin-related, cell cycle/S-phase genes, Rec8)
Fluorescence microscopy / GFP-dot sister chromosome cohesion assay S. cerevisiae P_SCC1-REC8, P_SCC1-SCC1, scc1Δ strains carrying P_GAL1-SCC1 and Chr5 tetO/GFP-TetR centromere label conditional SCC1 repression (glucose), benomyl mitotic arrest fraction of cells with one vs two GFP dots (cohesion loss)
Flow cytometry DNA content / cell cycle profiling S. cerevisiae Scc1, Rec8-expressing, and mad2Δ Rec8-expressing strains with P_GAL1-SCC1 SCC1 repression; G1 synchronization and release; mad2Δ DNA content (1C/2C) over time, S-phase and mitosis progression
Western blot protein level and stability S. cerevisiae strains with 3xHA-tagged Scc1 or Rec8 (P_SCC1 promoter), incl. esp1-1 mutant SCC1 repression; benomyl arrest; cycloheximide chase; esp1-1 temperature-sensitive separase kleisin protein level, cleavage product, half-life anti-HA antibody; Hxk1 loading control
Chromatin immunoprecipitation (ChIP) S. cerevisiae wild type, Rec8-expressing, untagged negative control G1 release, benomyl mitotic arrest chromosome-bound kleisin (DNA associated with immunoprecipitated kleisin) at cohesin binding sites anti-HA antibody
Sister kinetochore biorientation assay Rec8-expressing S. cerevisiae Rec8 substitution frequency of correct sister kinetochore orientation
Key results
  • Fitness of Rec8-expressing cells is 55% of wild type (45% reduction) 55% of wild type
  • Evolution for 1,750 generations across 15 parallel populations substantially increased fitness
  • Only one of the evolved populations had a mutation in Rec8; recurrent mutations occurred in mediator complex, cohesin-related genes, and S-phase-inducing cell cycle regulators 1 population with Rec8 mutation
  • In Rec8-expressing cells, fraction with two GFP dots rose from 10% in S phase to 50% in mitosis, indicating loss of cohesion 10% to 50%
  • Rec8 protein level during G1/S phase is lower than Scc1 4-fold less than Scc1
  • Rec8 half-life is shorter than Scc1; esp1-1 separase mutation restores Rec8 stability Rec8 58 min vs Scc1 200 min; esp1-1 raises Rec8 to 190 min
  • Removing Mad2 from Rec8-expressing cells increased the fraction of 1C cells, relieving the mitotic delay
  • Rec8-expressing cells showed more 2C-content cells at 30 min, consistent with faster/earlier S phase progression
Key statistics
  • other 45% fitness reduction (fitness 55% of wild type) (competitive growth, three biological replicates, two-tailed t test p<0.01)
  • count 1,750 generations (experimental evolution across 15 parallel populations)
  • fold_change 4-fold (less Rec8 than Scc1 during G1 and S phase)
  • other half-life: Scc1 200 min, Rec8 58 min, Rec8 in esp1-1 190 min (protein stability by cycloheximide chase in benomyl-arrested cells)
  • other 10% (S phase) to 50% (mitosis) (fraction of Rec8-expressing cells with two GFP dots; ≥100 cells per timepoint, 3 biological repeats)
  • pvalue p < 0.01 (two-tailed Student t test for fitness, protein level, and half-life comparisons)

Statistical methods review

Model: sonnet

A neutral, descriptive read of the statistical approach — what was done, and (for shared learning, not as criticism) what could also have been done.

The paper uses an experimental-evolution design in budding yeast, expressing the meiotic kleisin Rec8 in place of mitotic Scc1 and evolving 15 parallel populations for 1,750 generations. The primary phenotypic comparisons (fitness, protein levels, protein stability, and sister-cohesion frequencies) are made between two-strain pairings using two-tailed Student's t-tests, with results expressed as means ± SD from three or more biological replicates. Statistical significance is indicated with a threshold annotation (**p < 0.01) rather than exact p-values.

Replicationbiological Sample size3 biological replicates stated for most quantitative assays; ≥100 cells per time point for the cohesion microscopy assay; 15 parallel evolved populations for the evolution experiment GroupsRec8-expressing (PSCC1-REC8) vs. Scc1-expressing wild type, plus scc1Δ and esp1-1 mutant controls Pairingunpaired Randomization/blindingnot stated DispersionSD Exact p-valuesno Effect sizesno Confidence intervalsno Multiplicity correctionnone stated
Statistical tests used
Test Applied to n Assumptions
two-tailed Student's t-test Fitness of Rec8-expressing strain relative to wild type (Fig 1B) 3 biological replicates not stated
two-tailed Student's t-test Rec8 vs. Scc1 protein levels across synchronous cell cycle time points (Fig 2A) 3 biological replicates not stated
two-tailed Student's t-test Protein stability / half-life comparison of Scc1 vs. Rec8, and effect of esp1-1 mutation (Fig 2B) ≥3 biological replicates not stated
not stated (visual/descriptive comparison of proportion with two GFP dots) Sister chromosome cohesion assay over time in mitosis (Fig 1C) ≥100 cells per time point, 3 biological repeats na
flow cytometry (descriptive cell-cycle profiles, no explicit test stated) S-phase and G2/M progression comparison across strains (Fig 1D) not stated na
Approaches that could also have been used
  • Proportions of cells with two vs. one GFP dot were compared descriptively across time points and strains (Fig 1C)
    Could also: A chi-square test or Fisher's exact test (for small expected counts) on the counts of cells in each cohesion category at each time point could also be used — Proportion data with a fixed denominator (≥100 cells) follow a binomial distribution; a test designed for count/proportion data would directly quantify whether the difference in proportions between strains exceeds chance variation and yield a p-value tied to the specific comparison
  • Two-tailed Student's t-tests were used with n = 3 biological replicates for continuous measurements such as protein levels and fitness
    Could also: A Mann-Whitney U (Wilcoxon rank-sum) test could also be used for pairwise comparisons with very small n — With only three replicates per group, the normality assumption of the t-test cannot be verified empirically; a non-parametric alternative makes no distributional assumption and is often recommended when n is too small to assess normality
  • Multiple independent t-tests were conducted across several figures and comparisons without a stated multiplicity correction
    Could also: A single ANOVA (or repeated-measures ANOVA for the time-course data) followed by a post-hoc correction such as Tukey HSD or Benjamini-Hochberg FDR could also be applied across the family of comparisons — Conducting many pairwise t-tests increases the nominal type-I error rate across the experiment; a unified model with a post-hoc correction would control the family-wise error rate or false discovery rate while still identifying which specific pairs differ
  • Results were summarized using the mean ± one standard deviation
    Could also: A 95% confidence interval around the mean could also be reported alongside or instead of SD — With n = 3, SD describes spread of the sample but a CI directly conveys uncertainty about the population mean and is often easier to interpret for readers assessing practical significance; some journals and reporting guidelines (e.g., Nature Methods) now recommend CIs over SD for small n
  • Statistical significance was reported only as a threshold annotation (**p < 0.01)
    Could also: Exact p-values (e.g., p = 0.003) and a standardized effect size such as Cohen's d could also be reported — Exact p-values let readers judge the strength of evidence more precisely and are required for meta-analyses; effect sizes convey the magnitude of difference independently of sample size, which is informative when n is very small
  • Cell-cycle progression was characterized by flow cytometry profiles displayed as visual traces at selected time points (Fig 1D) without a stated statistical test
    Could also: A quantitative summary such as the mean time to reach 50% of cells in G2/M (T50) with a t-test or Kaplan-Meier–style analysis of cell-cycle exit could also be applied — Reducing continuous time-course data to a single derived quantity (e.g., T50 or area under the curve) enables formal statistical comparison across strains and replicates, complementing the visual impression from gel-lane profiles
Software: not stated in provided text

Result convergence & founder nodes

Findings this paper shares with others that ran a comparable experiment. A node’s strength is how many independent papers report it (replication breadth) — not how often it is cited, so a heavily-replicated but under-cited founder still stands out.

Citation network

Where this publication sits in the reproducibility-weighted citation graph — what it is built on, and what is built on it. Citation data from OpenAlex.

Citations
24
Impact: medium
Foundation confidence
None of its references are in our reproducibility record yet — its foundation cannot be assessed.
Topics

No assessed neighbours yet — the network grows as more papers are assessed.

Data lineage

The datasets this paper uses (text-mined from the full text via Europe PMC), and which other assessed papers stand on the same data. A shared dataset is a factual link — not a judgement.

PRJNA594153 BioProject in Methods (http://purl.org/orb/Methods)
no other assessed paper uses this yet

What was reproduced

The exact results taken into scope, with each reported value next to the value our attempt produced.

Scope — pmid-32155147

Paper: Hsieh YP, Makrantoni V, Robertson D, Marston AL, Murray AW (2020). Evolutionary repair: Changes in multiple functional modules allow meiotic cohesin to support mitosis. PLoS Biol 18(3):e3000635.

System: S. cerevisiae in which the only cohesin kleisin is the meiotic Rec8 (mitotic Scc1/Mcd1 deleted). Founder is sick; 15 populations were evolved (~1,750 generations) and recovered fitness. The paper asks which mutations drive the adaptation.

Pipeline-derived results (IN SCOPE)

The central computational result is whole-genome variant calling of the evolved populations against the ancestor, yielding the list of adaptive mutations (non-synonymous, frequency ≥90%) reported in:

  • S1 Table — mutations at generation 375
  • S2 Table — mutations at generation 1,750
  • Summarised in Fig 3B/3C (functional modules hit).

Reported high-level findings to check against:

  • Cdk8 / mediator module (SSN2/MED13, SSN3/CDK8, SSN8/CYC8?, SRB8/MED12): hit in 12 of 15 populations; 9 produced early stop codons.
  • Cohesin: SMC3 in 7 populations, SMC1 in 2, ESP1 in 4.
  • G1→S regulators (MBP1, CLN2, SWI4, SWI6): 6 populations total.
  • Chr IX (chromosome 9) aneuploidy in several populations.
  • One population (P12) carries REC8 E148Q on the plasmid copy.

Pipeline / methods named

  • Methods: "WGS data were processed as described [ref 61]".
  • Cited analysis code: github.com/koschwanez/mutantanalysis (J. Koschwanez). Pipeline = bwa (aln/samse/sampe) → samtools sort/index → GATK IndelRealigner → VarScan (variant calling, ancestor vs sample) → mutantanalysis.py classifies SNP/indel as synonymous / non-syn / stop and reports per gene. Reference genome shipped in repo: S. cerevisiae S288C (SGD) with SGD_features annotation.

Data

  • SRA BioProject PRJNA594153 (GEO GSE141598). 120 paired-end runs:
    • E0-xx (5 ancestral clones: 03,05,09,11,13) — the founders.
    • E3-P1..P15 (15 populations, earlier timepoint).
    • E14-P1..P15 (15 populations, later timepoint, ~gen 1,750).
    • YPH.../YPH-II... — backcross segregant pools (out of primary scope).

Reproduction strategy (80/20)

Map the koschwanez tool onto the paper's pooled-population data: treat an E0 ancestral clone as ancestor and an evolved population pool as clone (segregation/pool args are optional in mutantanalysis.py). For a few CLEAR data points:

  1. Align ancestor + selected evolved pools to S288C with bwa + samtools.
  2. Call variants with VarScan (the tool's own caller) — sample vs ancestor.
  3. Annotate (snpEff S288C R64 / SGD features) → non-syn, stop, frequency.
  4. Filter to non-syn ≥90% and check whether the specific genes/mutations the paper reports for those populations are recovered (gene, type, AA, freq).

Faithfulness note: the tool is Python2 + bwa aln + GATK3 + legacy samtools (env-fragile). The substance — VarScan calling against the ancestor — is reproduced with current builds of the same caller; the GATK3 indel-realign step (deprecated, optional refinement) and the HTML classifier are not required to recover the reported mutations. This is recorded honestly, not as a 1:1 rerun of the exact legacy binaries.

OUT OF SCOPE (not attempted)

  • Wet-lab phenotypes (growth/fitness assays, microscopy, tetrad dissection).
  • ChIP-seq analysis (separate repo PhoebeHsieh-yuying; different result class).
  • Backcross-segregant linkage (YPH pools) — secondary, large, optional 20%.
  • Exact reproduction of legacy GATK3 indel realignment / HTML output.
Figures / tables: Table
P15_SMC1_VI_120680
Reported
chr6:120680 A>C 100% (cohesin SMC1, nonsyn)
Reproduced
VI:120681 A>C 100% SS=2 SMC1/missense_variant
exact
P15_ESP1_VII_683739
Reported
chr7:683739 A>T 95% (separase ESP1, nonsyn)
Reproduced
VII:683740 A>T 95% SS=2 ESP1/missense_variant
exact
P15_SSN2_IV_1346880
Reported
chr4:1346880 T>A 100% (Cdk8/mediator SSN2, nonsyn)
Reproduced
IV:1346881 T>A 100% SS=2 SSN2/stop_gained
exact
P13_SMC3_X_299226
Reported
chr10:299226 G>A 100% (cohesin SMC3, nonsyn)
Reproduced
X:299227 G>A 100% SS=2 SMC3/missense_variant
exact
P13_SRB8_III_257464
Reported
chr3:257464 C>A 95% (Cdk8 SRB8, nonsyn)
Reproduced
III:257465 C>A 95.35% SS=2 SRB8/stop_gained
exact
P10_SRB8_III_258608
Reported
chr3:258608 C>G 100% (Cdk8 SRB8, nonsyn)
Reproduced
III:258609 C>G 100% SS=2 SRB8/missense_variant
exact
P10_BOI2_V_390824
Reported
chr5:390824 G>C 100% (BOI2, nonsyn)
Reproduced
V:390825 G>C 100% SS=2 BOI2/stop_gained
exact
P4_SSN3_XVI_474599
Reported
chr16:474599 indel 96% (Cdk8 SSN3, nonsyn frameshift)
Reproduced
XVI:474600 GT>G 93.18% SS=2 SSN3/frameshift_variant
within tolerance
ALL_21_reported_SNPs_4pops
Reported
21 nonsyn SNPs across P4,P10,P13,P15 (S-Table Gen375)
Reproduced
21/21 recovered as high-conf VarScan SOMATIC with identical ref>alt and matching frequency
exact
TIR1_V_175658_recurrent_indel
Reported
TIR1 chrV:175658 indel at 91-96% in 6/15 populations
Reproduced
-36bp deletion present at only ~5-7% in BOTH ancestor and evolved; not a high-conf somatic call (GATK indel-realign omitted)
did not match

Assessments & scoring basis

Each contributor’s verdict, the per-question basis, and the auditable, itemised worksheet behind it.

🤖 AI curator · claude (ai-curator room) · v1.0 L1 90/100

An automated assessment. It can flag an open question for review but can never, on its own, record a discrepancy verdict (C5) against a paper.

🟢1. Data identity
🟢2. Endpoint comparability
🟡3. Location of the main deviation
🟡4. Cause of the deviation
🟡5. Derivability / plausibility
🟡6. Severity of the deviation
🟢7. Core claim
🟡8. Severity of the miss (overall human judgment)
Scoring basis — itemised

Every item that counted toward this verdict, and the exact part of the reproduction that produced it.

Supporting (toward a concern)
Content-critical question only partially held
+2 pts
From: Q5 · Derivability / plausibility 🟡
Content-critical question only partially held
+2 pts
From: Q8 · Severity of the miss (overall human judgment) 🟡
Minor / cosmetic deviation
+1 pts
From: Q3 · Location of the main deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q4 · Cause of the deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q6 · Severity of the deviation 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +5

Strong reproduction. All 21 reported non-synonymous SNPs across the four sampled gen-375 populations were recovered from the raw SRA reads with identical ref>alt and matching allele frequency, including every headline module mutation (cohesin SMC1/SMC3, separase ESP1, and Cdk8/mediator SSN2/SRB8/BOI2 confirmed as stop_gained), so the paper's central conclusion holds 1:1. The only systematic difference is a cosmetic +1 coordinate offset (0-based paper vs 1-based VCF). One genuine flag remains for human review: the recurrent TIR1 chrV:175658 indel reported at 91-96% in 6/15 populations is only ~5-7% in the raw reads of both ancestor and evolved pool — plausibly an artifact of the paper's omitted GATK indel-realignment, but not derivable from the deposited data and biologically suspicious as an identical-position recurrent indel. Overall a solid reproduction with explainable deviations.

🤝
Reproduced automatically — and fairly

Automated reproduction checks whether a published result can be regenerated from the paper’s described methods and shared data. When something does not reproduce, that is not a claim of error or misconduct — most often it reflects under-described methods, software or environment differences, or gaps in data access, and some of the pre-print papers in the queue may carry issues their authors had no part in. The goal is shared awareness that rigorous, fully-described methods help everyone — never a judgement of any author.

Are you an author? We would genuinely like to hear from you — to clarify the record, add data or code, re-run the pipeline after an accession update, and publish your response right next to the assessment. Everything here is open and auditable.

🚩 Report an error in this record

Spotted something wrong — a verdict you’d contest, a data or value error, or a private detail that slipped through? Tell us, with a short justification. Authors and readers are equally welcome to write in; we review every report.

Prefer email, or the form below not working? Contact us at support@doesitreproduce.com.

Reproduction footprint

claude-opus-4-8

Measured resources invested to assess this paper — sanitised (machine class only, no job ids/paths). Compute = HPC accounting (SLURM); tokens = the AI agent's session.

381.6 k
tokens (I/O) · 68.6 M incl. cache
43 min
runtime · 0.72 CPU-h
3.8 GB
peak RAM
4
HPC jobs
hummel
machine