Issues

• A spending cap can only be set on the command line: not in the builder, on a host, per unit, or shown on the job page

  enhancement
  Feature

  ThinkFlowLab/nullius#45 · Filter by author aitianshi opened on Nov 10, 2026
  

• A test helper races with the process it watches and fails under load

  bug
  Bug

  ThinkFlowLab/nullius#40 · Filter by author aitianshi opened on Nov 10, 2026
  

• Model prices are a table in the code: OpenAI only, and fixed between releases

  enhancement
  Feature

  ThinkFlowLab/nullius#38 · Filter by author aitianshi opened on Nov 10, 2026
  

• SWE-agent cannot be run by the standard as shipped: it needs root in the kits and the internet at start

  enhancement
  Feature

  ThinkFlowLab/nullius#37 · Filter by author aitianshi opened on Nov 10, 2026
  

• LongMemEval: offer the upstream's index expansion from the authors' released caches

  enhancement
  Feature

  ThinkFlowLab/nullius#36 · Filter by author aitianshi opened on Nov 10, 2026
  

• SWE-bench Verified: a trial that changed nothing is graded as a patch, and resolves django__django-10097

  bug
  Bug

  ThinkFlowLab/nullius#33 · Filter by author aitianshi opened on Nov 10, 2026
  

• No way to tell when a pinned image has been rebuilt upstream

  enhancement
  Feature

  ThinkFlowLab/nullius#31 · Filter by author aitianshi opened on Nov 10, 2026
  

• τ³-bench: offer the reranker and qwen search methods through the relay

  enhancement
  Feature

  ThinkFlowLab/nullius#28 · Filter by author aitianshi opened on Nov 10, 2026
  

• A run keeps its pace into a rate limit and never runs a rate-limited unit again

  enhancement
  Feature

  ThinkFlowLab/nullius#27 · Filter by author aitianshi opened on Nov 10, 2026
  

• Let a benchmark say which tasks need the internet, so Terminal-Bench's 36 left-out tasks can run

  enhancement
  Feature

  ThinkFlowLab/nullius#26 · Filter by author aitianshi opened on Nov 10, 2026
  

• SWE-bench Verified: agents start outside the instance's conda environment and cannot import the project

  bug
  Bug

  ThinkFlowLab/nullius#24 · Filter by author aitianshi opened on Nov 10, 2026
  

• A machine requirement cannot say how much GPU memory it needs, so Stella starts on a card too small

  enhancement
  Feature

  ThinkFlowLab/nullius#23 · Filter by author aitianshi opened on Nov 10, 2026
  

• Three limitation lines that are an option, a dangling reference and a test fixture

  enhancement
  Feature

  ThinkFlowLab/nullius#22 · Filter by author aitianshi opened on Nov 10, 2026
  

• Terminal-Bench tasks have no tmux: git-multibranch's reference solution and Terminus 2 both fail

  bug
  Bug

  ThinkFlowLab/nullius#21 · Filter by author aitianshi opened on Nov 10, 2026
  

• HumanEval: say which helper files of the workspace the check does not run

  enhancement
  Feature

  ThinkFlowLab/nullius#20 · Filter by author aitianshi opened on Nov 10, 2026
  

• HumanEval's Original setup runs model-written code on the machine without asking

  bug
  Bug

  ThinkFlowLab/nullius#19 · Filter by author aitianshi opened on Nov 10, 2026
  

• Say three verified upstream defects in the benchmarks' upstream limitations

  enhancement
  Feature

  ThinkFlowLab/nullius#18 · Filter by author aitianshi opened on Nov 10, 2026
  

• Stop a run at a spending cap

  ThinkFlowLab/nullius#14 · Filter by author ttachon opened on Nov 10, 2026
  2

• The mem0 capability never gives Mem0 the session's date, so its facts carry the day of the run

  bug
  ThinkFlowLab/nullius#12 · Filter by author ttachon opened on Nov 10, 2026
  2