# Brief — Agent E (AI Research Specialist), Phase 5

Scope: large language models, foundation models, multimodal AI, generative AI, AI agents, reinforcement learning, machine-learning theory, deep-learning architectures, AI safety and alignment. Area directories: ai_ml. Manuscript target: Part V — The Artificial Intelligence Transformation (file manuscript/sections/part05_ai.tex — you write the disciplinary synthesis there; for Part VII the two specialists write separate files part07a_theory.tex / part07b_security.tex that the director merges). Follow agents/briefs/SPECIALIST_REVIEW_TEMPLATE.md and agents/briefs/WRITER_COMMON.md. Inputs: literature_synthesis/landmarks/landmarks.csv and abstracts.jsonl for your areas; results/graph/category_counts_yearly.csv, category_flows_yearly.csv, results/bibliometrics/cagr_by_category.csv, results/econometrics/breaks.csv (rows cat), results/emerging/*.csv, data/processed/graph/extraction/extractions*.jsonl (knowledge-graph extractions with provenance, if present), manuscript/references_all.bib. Deliverables: literature_synthesis/<agent_dir>/{claims.csv, contributions.csv, review.md, open_problems.csv} AND the LaTeX section (20–35 pages) that turns review.md into manuscript prose with \citep keys (verify new ones), figures/tables referenced by label (do not create figures; list the ones you need in your report), numbers only through macros of manuscript/tables/numbers.tex or by citing the result file in a footnote when no macro exists. Distinguish established / contested / speculative; never call a preprint peer-reviewed; abstracts are not full texts. Budget: ≤ 25 lines report at the end.