How I AIHow this former NYT columnist uses ChatGPT to brainstorm, do research, and find the perfect metaphor
CHAPTERS
- 0:00 – 2:41
Replacing Google with ChatGPT for faster idea research
Claire asks how Farhad brainstorms without defaulting to Google. Farhad explains how ChatGPT with web search can surface key players, context, and reading links in minutes—work that used to take hours.
- •Uses ChatGPT web search as a starting point for a new topic
- •Quickly identifies major stakeholders and prevailing commentary
- •Provides links for deeper reading instead of manual searching
- •Compresses hours of initial research into a short interactive session
- 2:41 – 4:20
From skepticism to “two windows open” in every writing session
Farhad recounts his early reaction to ChatGPT’s first releases and why he didn’t find them useful for writing at first. As quality improved, it became a constant companion in his workflow, always open alongside his draft.
- •Early models produced weak prose but improved quickly
- •Adopts tools early and looks for practical leverage, not replacement
- •First major win: word-finding and nuanced synonym selection
- •Now writes with ChatGPT always open next to his document
- 4:20 – 6:54
Live demo: web-search prompting to map the debate on a topic
Farhad walks through a concrete example prompt (commentary around Trump’s tariffs) and shows how he interrogates the model for specific angles and industry viewpoints. The emphasis is on conversational drilling rather than one-shot queries.
- •Turns on web search and asks for supportive viewpoints and notable commentary
- •Uses follow-up questions to narrow to sectors (e.g., automotive)
- •Treats the model like an interactive research assistant
- •Moves from general landscape to targeted sub-questions quickly
- 6:54 – 8:35
How he evaluates AI-sourced information and avoids hallucinations
Claire challenges the reliability of AI research. Farhad explains how citations and source lists changed the usefulness of the tool and describes a lightweight verification habit—clicking through when something seems off.
- •Relies on linked citations next to claims for accountability
- •Checks the full list of consulted sources when needed
- •Notes earlier versions were risky due to opaque sourcing and hallucinations
- •Uses verification to go deeper faster without blind trust
- 8:35 – 10:53
Finding new angles: using AI to propose structure and arguments
Farhad describes how ChatGPT reduces the hardest part of writing: figuring out where to start. He uses it to surface compelling frames, main points, and overlooked perspectives, then debates and refines them through back-and-forth.
- •Asks for compelling arguments, main points, and what to highlight
- •Uses AI suggestions to uncover ideas he might not have considered
- •Compares it to a capable research assistant (not a full editor)
- •Benefits from low-friction iteration—no social cost to rejecting ideas
- 10:53 – 14:31
Metaphor and idiom workshop: escaping clichés like “pay the piper”
Farhad demonstrates how he replaces tired expressions by pasting a draft sentence and asking for metaphorical alternatives. The conversation becomes a mini language lab—evaluating nuance, correctness, and coherence of imagery.
- •Flags clichés and searches for fresher idioms and metaphors
- •Uses AI to find idiom-like alternatives that traditional thesauruses miss
- •Iterates by critiquing suggestions and requesting fixes to imagery
- •Uses AI to explore origins/nuance of expressions and choose the best fit
- 14:31 – 17:09
Remote-era “newsroom substitute”: ChatGPT as an always-available partner
Claire connects the workflow to the loss of live newsroom banter. Farhad explains the similarity to Slack-style collaboration: it doesn’t replace humans, but it can approximate the bouncing-of-ideas function with instant availability.
- •Chat interface mimics Slack conversations with colleagues
- •Acknowledges it’s not human, but functionally similar for idea-bouncing
- •Uses it to refine and correct metaphor coherence in real time
- •Addresses creator fears: tool accelerates craft rather than replacing voice
- 17:09 – 20:13
Word-level precision: the ‘super-thesaurus’ approach
Farhad shows how he uses AI for rapid synonym exploration and nuance checks while drafting. Claire highlights that the model can categorize options by tone (dramatic, colloquial, ironic), which helps match intent to word choice.
- •Uses AI constantly to replace or sharpen single words (e.g., outrage)
- •Faster and more flexible than searching a traditional thesaurus
- •Can ask whether a word is correct and discuss shades of meaning
- •Explores tone-based groupings of synonyms to match rhetorical intent
- 20:13 – 22:32
AI as a critical first reader for early draft direction and structure
Farhad explains his practice of handing ChatGPT partial drafts (six or seven paragraphs) to test whether the piece is landing. He focuses on structure, clarity, and speed to argument rather than expecting deep logical validation.
- •Shares partial drafts to evaluate direction before a full editor review
- •Asks if the point is clear quickly enough and where to tighten
- •Uses feedback to reduce unnecessary commentary and improve phrasing
- •Iterates: write → get input → revise → continue writing → polish
- 22:32 – 25:04
Lightning round: what he wants next and how he handles bad outputs
Farhad names product gaps that would make the workflow smoother—persistent memory and better cross-app context without copy/paste. He also shares his strategy when the model goes off track: be blunt, and stop using it when it’s unhelpful.
- •Wants stronger persistent memory across chats and time
- •Wants broader ‘see my screen’/cross-app context to reduce copying
- •Uses direct, brusque corrections to redirect the model
- •Recognizes limits and disengages when conversations become circular
- 25:04 – 25:45
Closing remarks and where to find the show
Claire thanks Farhad for revealing a concrete, non-generic way AI can strengthen writing. The episode ends with a standard call to like, subscribe, comment, and find the podcast on major platforms.
- •Recap: AI can make writing more specific and impactful (not generic)
- •Appreciation for transparent behind-the-scenes workflow
- •Call to like/subscribe/comment and leave ratings/reviews
- •Links viewers to the show website and podcast platforms
