Kiloloop Research

Curated research from Kiloloop — runtime comparisons, operational reviews, practical guides, and protocol patterns for multi-agent coordination.

What's here

Research published from our internal pipeline. Each piece is reviewed and sanitized before landing here.

Directory	Contents
`runtime-comparison/`	Runtime capability comparisons and benchmarks
`ops-reviews/`	Operational reviews of our own agent fleet, with the scripts behind the numbers

Date	Piece	Where
2026-09-10	Validated dispatches, defined: what the number on our homepage counts, and what it leaves out — the definition behind the dispatch count: four tests, the exclusions, the count over time with its correction to 557 dispatched legs, what the count is made of, and failure and human-intervention criteria from both records; aggregates, a redacted sample and `CITATION.cff` ship alongside, and the numbers come from `estimate_stats.py` and `hil_stats.py`	`ops-reviews/`
2026-09-07	Estimate accuracy, measured: seven months of dispatch estimates against their actuals — actual-to-estimate ratios from 606 dispatch rows by period, month, estimate size and task class, the declared-envelope headroom on review-loop legs, verdicts and review status; the computation behind the numbers is `estimate_stats.py`	`ops-reviews/`
2026-09-05	Human in the loop, measured: four months of autonomy-gate records — pause rates, human outcomes and latency, checkpoint breaches and declared-vs-actual fidelity from 602 audit records; reproducible with `hil_stats.py`	`ops-reviews/`
refreshed 2026-09	Runtime capability matrix — what each coding-agent runtime can and cannot do, self-reported and cross-verified by the agents	`runtime-comparison/`
refreshed 2026-09	Prompt caching patterns — Claude and Codex caching defaults, billing mechanics, and measurement	`runtime-comparison/`