Hacks/Hackers × The New York Times
// program

AI × Investigative Journalism Forum

Friday, September 25, 2026_

The New York Times / 620 Eighth Ave / 15th floor

The Forum / main The Sandbox The Idea Lab

Three tracks run in parallel late morning and afternoon — pick one per slot. Cards are color-coded by room.

Taking notes? Everyone is writing in one shared doc, with a tab for each session.

// morning
9:00 AM30 min
Registration & breakfast
9:30 AM15 min
The Forum

Welcome & games

Burt Herman / Paul Cheung (Hacks/Hackers) / Rubina Fillion (NYT)

9:45 AM60 min
The Forum

An illustrated guide to A.I. for investigative reporting

Dylan Freedman (NYT) / Juliana Castro Varon (NYT)

Details

A hand-drawn presentation on the art of applying A.I. to investigative reporting from two multidisciplinary journalists on The New York Times's A.I. Initiatives team. In this session, they will share how A.I. is and isn't useful to the journalistic process, emphasizing the importance of human input and expertise. From needle-in-the-haystack multimedia analyses to holding the powerful A.I. companies that shape our everyday lives to account, they will share stories, tools and best practices they've uncovered as they've taught journalists to use A.I. responsibly.

10:45 AM15 min
Break
11:00 AM60 min / pick one
The Forum

When errors are expensive: A.I. evaluation for investigative journalism

Teresa Mondria Terol (NPR) / Bruno Savoca Albors (JPMorgan)

Details

A reporter can now have a model classify 250,000 public comments in an afternoon. The moment the story says 63% of them oppose a rule, the newsroom owns a harder question: how do we know that number is right? Legal and finance teams answer it with rubrics and sampling plans that penalize unsupported claims. This session brings those methods across, and covers what evaluation failures reveal about where AI should not be used at all. The examples come from work including The Visa Pulse, an automated system that tracks U.S. immigration policy.

The Sandbox

Bringing journalism out of the dark ages

Philip Bump (Hearst)

Details

The facts that explain a community sit in public records, agency portals and a newsroom's own back catalog of reporting, in formats that resist a quick query at exactly the moment a reporter is on deadline. Hearst Connecticut's data team has been building AI tooling that lets reporters pull answers grounded in those records and turn them into repeatable workflows. An honest account of building it inside a working newsroom: what the effort is worth, where it fails, and how to keep a human between the analysis and the reader.

The Idea Lab

Finding the headline in a mountain of documents: Hands-on with an AI-powered investigation tool

Chuck Kellner / Chris Miles / Sean Wiederkehr / Renata BystritskyEverlaw

Details

Try out the tools that Le Monde, USA Today, the Associated Press and WIRED use in large document investigations to find the facts that matter. Everlaw is an AI-powered document review platform used by legal teams to identify evidence and build case narratives verifiable in the courtroom, and it's available to journalists at free and heavily discounted rates. You'll run a realistic investigation in a live sandbox: ask questions across 100k documents and get back answers with excerpts, source links, and a fact-check trail your editors can replicate. Bring your laptop, and sign up in advance (by Sept. 23) to get access to the sandbox.

12:00 PM75 min
Lunch
// afternoon
1:15 PM60 min / pick one
The Forum

Revealing the truth, avoiding harm: Ethical practices for A.I. in investigations

Jon Greenberg (Poynter) / Sisi Wei (CalMatters) / James O'Toole (NYT)

Details

A working framework for spotting the ethical risks that different kinds of investigations carry, and the practical steps that shrink them. The session covers validating AI output drawn from large datasets, using complementary methods to push accuracy up, enlisting public feedback to improve data quality, and the particular trouble that arrives when AI makes qualitative rather than quantitative judgments.

The Sandbox

Vibe-coding for open-source investigations

Avery Schmitz (CNN)

Details

Flight data, satellite imagery, vessel tracking: the hard part of an open-source investigation is rarely any single tool, it's getting messy sources to work together. This is a build session, going from a vague hunch to a working script, then showing how a throwaway flight-tracking API call at CNN grew into an automated analysis pipeline. It closes on the safeguards that keep AI-generated code from becoming a published mistake. Attendees will work with live flight data. Bring a laptop and an AI account, setup details to follow.

The Idea Lab

Throw out your RAGs: Using custom tools to report with document dumps

Allison Martell (Reuters)

Details

Two years of experiments on Reuters' data team point one way: preprocessing documents and pointing an LLM at custom-built search tools beats RAG and semantic search almost every time. The session works through what did and didn't hold up, from a 2024 famine investigation to Syrian government documents to a recent story on Iran's largest cryptocurrency exchange, plus quirkier experiments like multimodal embeddings of handwritten pages. You'll leave with a pattern for finding the news in a document set too big to read in full.

2:15 PM15 min
Break
2:30 PM60 min / pick one
The Forum

Digging into the Epstein Files

Duy Nguyen (NYT) / Larry Fenn (AP) / Kriti Singh (NPR)

Details

Journalists and technologists from three newsrooms will share case studies of how they used A.I. to sift through millions of documents in The Epstein Files. They’ll explain how they surfaced connections and identified leads worth reporting, while preserving editorial judgment and verification. This session offers a behind-the-scenes look at the process of rapidly prototyping tools and workflows for a major developing story with a vast archive.

The Sandbox

Building a verification toolkit for investigative journalists

Zoe Darme (Google DeepMind) / Zeve Sanderson (Google)

Details

In this hands-on workshop, you’ll be introduced to Backstory, which helps uncover the history and context of online images, and SynthID, which helps identify AI-generated content. Participants will then tackle a real-world verification challenge and step into the role of product manager. Working in small groups, you’ll identify the evidence reporters need, where AI helps, where human judgment matters, and the guardrails investigative tools require. By the end, each group will design a verification workflow or product concept grounded in a real investigative reporting need.

The Idea Lab

Build your own A.I.-assisted backgrounding workflow

Adiel Kaplan (CUNY / Tow-Knight Center)

Details

Backgrounding a company or a person eats the early weeks of an investigation, and it is the same set of moves every time: the same sites, the same databases, the same checks. Repetition like that is worth automating. You'll build a skill that runs your process for you and returns a dossier with what it found, where it searched and auditable sourcing. The method is written up in After the Moat. Bring a laptop and an AI account, setup details to follow.

3:30 PM15 min
Break
3:45 PM50 min
The Forum

Lightning talks

4:35 PM10 min
The Forum

Closing

// after the forum
5:00 PM5:00–7:00 PM

Happy Hour / hosted by The New York Times

Printers Alley / 215 W 40th St / about a block from the Times

Pre-registered guests only — space is limited

venue: The New York Times, 620 Eighth Ave
floor: 15 (sessions) · reception offsite
entry: pre-registered Times visitors only — bring photo ID
notes: shared notes doc · a tab per session