I heard the vocabulary of this year’s New York Tech Week — surface, alignment, scale, friction, trust — circulate through nearly every conversation. These words accrued authority through repetition. Sounding incisive and significant, but elastic enough to mean anything, they describe real, consequential activity in one room but serve aspirational ideals in another. It’s the air we breathe. Of all the terms tossed around, trust was deployed the most but defined the least.
I attended seventeen events in six days at New York Tech Week (June 1–6, 2026), my first time at this decentralized, self-organizing, and occasionally self-combusting conference. I’m a writer and editor coming from the nonprofit, museum, and scholarly publishing worlds. I went to listen, and to identify the gaps. I learned a lot.
People talked about clean data and C-suite buy-in. Design agencies are rethinking billing hours now that everyone can vibe code. No one has cracked the SEO/AEO code — what Michael King of iPullRank called “relevance engineering.” At a Fashion Institute of Technology panel, Diarra Bousso described market testing dead stock fabric on Instagram to get instant feedback on what’s popular right now. She designs and ships new styles made from leftover material right away, keeping her business sustainable while eliminating waste. At Pier 57 I learned about Show Your Work Lab, an open-source project for encrypting metadata in digital photography, cofounded by the Norwegian photojournalist Nora Savosnick. If widely adopted, it’s a terrific way to authenticate images and dispute deepfakes.
Sustained context management for LLMs is notoriously shaky, a problem not yet solved. A meetup on “Agents in Production: The Memory Problem,” hosted by Monday.com, featured engineers trying to solve this drift. Jameson Lee of Pydantic AI advocated directing and improving agents by introducing memory, observability, and runtime. Tobie Morgan Hitchcock described Spectron, a new SurrealDB system that logs where a new fact comes from and how it was confirmed before distributing it system wide, across knowledge graphs, vector stores, and relational databases. New information supersedes earlier versions, but the correction history is maintained, not overwritten. Evan Rimer of Tavily stressed the credibility of LLM citations in real time from the end user’s point of view.
The presentations periodically got murky for a generalist, but I understood these engineers are grappling with archival science. As someone close to the disciplines that develop provenance and authority, I felt at home. Librarians have been engaged for years with the issues stumping developers now; they may have unexpected solutions. A Google engineer spoke from the floor, asking the speakers if their products analyze user prompts to teach us how to ask questions better while reducing token usage. Replies were noncommittal, but the idea’s worth thinking about.
AI agents are now managers, leading teams of subordinate agents busy with their own tasks. How do we supervise the supervisors, and can we trust their work? At “AI Agents in the Wild,” organized by PBD TokenRouter and Ezio AI, Bryant McCombs of OpenAI defined trust as something credible, reliable, and intimate. That third term was surprising to hear — I’d like to know how intimacy can scale. As companies codify AI workflows and protect proprietary data, Google’s Diane Chang recommended that AI governance be its own full-time job. McCombs envisions AI Resources Departments catching on, because judgment, accountability, and visibility, he said, are not going away. Mario Giacalone of BlackRock agreed, adding that solid governance can result in better usage and higher profits.
Museums and galleries are engines for real estate and tourism. They’re also staffed with professionals whose jobs are to evaluate, classify, and present information with confidence. Those people weren’t in the room. I return to SurrealDB’s demo and wonder, “What must a record carry before it deserves belief?” The trusted institutions that have answers — libraries, archives, higher education — weren’t well represented at New York Tech Week.
Many working in the arts and humanities recognize that AI is fundamentally reshaping not just how we conduct research, but how we express ourselves. Some fear AI will replace them entirely. Historians, philosophers, architects, designers, curators — they have smart and necessary perspectives to add, and all are integral to our shared future. One encouraging idea ran through my mind as I shuttled from panel to panel: people who know language intimately will soon be in demand.