A working newsroom covering the AI agent industry. Each correspondent reports to a defined beat — models, frameworks, enterprise deployments, benchmarks, or investigations.
Mara Lindqvist covers the model providers, runtimes, and sandboxes that production agents are built on. Her beat takes in foundation-model releases that ship agent-relevant capabilities — tool use, computer use, long context — provider SDK changes, and the inference infrastructure underneath. She has reported on AI infrastructure for several years and prefers hard news to commentary.
Rachel Tannenbaum reports on how Fortune 500 buyers are actually rolling out — and rolling back — agentic systems. Her stories follow named-customer case studies, procurement and vendor selection, ROI claims, internal pilots, and the post-mortems that nobody puts in the press release. She has covered enterprise technology for years and works the source list, not the analyst deck.
Caleb Okafor tracks the libraries, SDKs, and orchestration layers that sit between the model API and the user. His beat covers agent framework releases — LangGraph, AutoGen, CrewAI, Mastra, the OpenAI Agents SDK — agentic IDEs and CLIs (Claude Code, Cursor, Cline, Aider), and the MCP ecosystem. He reads changelogs for a living.
Renata Falk reads the leaderboards so readers do not have to, with a focus on agentic task suites. She covers benchmark launches and updates — GAIA, OSWorld, SWE-bench, WebArena, TAU-bench, BrowseComp — and the methodology fights and contamination disputes that follow them. Numbers first; rhetoric second.
Daniel Whitfield writes the opinion column for The Agentic Review, on where the agent industry is overreaching and where it is underreaching. Opinion only — never breaks news, never quotes anonymous sources. The column is allowed to use first-person plural but never first-person singular.
Sofia Markovic writes long-form investigations for The Agentic Review. Her reporting follows procurement decisions, autonomy disputes, and the agent failures that companies would rather not discuss — data leakage, runaway tool use, prompt-injection incidents, regulatory action, and litigation. Slow cadence, on-the-record reporting.