Podcast
Product Fridays
Hosted by The Andrews · EN
Andrew McGill and Andrew Phelps talk creativity, emerging tech, and the art of building great products. Subscribe for weekly inspiration and insights.
productfridays.substack.com
productfridays.substack.com
12episodes
Episodes
Newest firstAll episodes
How Can News Publishers Compete with AI Digests? We See Two Ways.
Oct 1300:21:45Tap to summarizeA programming note: We’re now releasing Product Fridays on Mondays. We know, we know, Friday is right there in the name. But Friday is the vibe, you see. And Monday turns out to be a better day to release podcasts.OpenAI’s latest product, ChatGPT Pulse, automatically compiles a personal news digest from your email, calendar, and across the web. It’s convenient, addictive — and a little terrifying if you work in media. Because in this world, the reporter still digs up the facts, but the reader never actually reads their words. The AI does the writing instead.This week on Product Fridays, we ask: What happens when the machines become the primary consumers of journalism?We draw a distinction between “bespoke” media — the creators and publications you seek out for their voices (The Atlantic, Substack, the McElroys) — and “commodity” news, where the value lies in facts, not style. The former might survive as personality-driven brands. The latter risks becoming raw material for AI.Or is there a third way? Consider a new model that lets publishers feed high-quality reporting directly to LLMs and get paid for it. Enter MCP, a universal protocol that allows language models to securely talk to other systems. Imagine paying a few dollars to add CNBC, The Economist, or the AP directly to your personalized feed.It’s a hopeful idea: reconnecting readers, machines, and journalists in a way that rewards everyone. But unless news organizations adapt fast, ChatGPT could do to media what Apple News, Google, and Facebook already did, only faster.And hey, we’ve got a request. We’re asking our readers/viewers/listeners to leave a rating and review on Apple Podcasts. As thanks, we will write a personalized limerick for every review — and read it on the next episode. Win-win-win-win-win-win-win!Mentioned in this episode:* Introducing ChatGPT Pulse* Model Context Protocol (MCP)* The McElroy Family This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit productfridays.substack.com
Transcribe →
ChatGPT Can Do Everything. That’s the Problem.
Sep 2600:18:54Tap to summarizeAndrew McGill shares a new demo he’s been working on — an AI-powered interior design assistant called Zhuzh — and wonders aloud: if this is just ChatGPT with a prettier interface, what’s the point? But maybe that is the point.We go deep on:* Why good UI still matters in the age of LLMs* How purpose-built tools outshine general-purpose chat for many use cases* Where wrapper apps land on the spectrum of AI product value* The difference between novelty and utility* What makes an AI idea ChatGPT-proofPlus, honorable mentions: personalized podcast feeds, offline LLMs, and the dream of an air-gapped house AI.We also commit to launching a small wrapper app of our own before the next episode. What should we make? Tell us in the comments.Mentioned in the episodeHere are the products and articles we talked about (links TBD):* Zhuzh, Andrew McGill’s AI-powered interior design demo* OpenAI usage study (700M weekly active users)* Flighty (flight tracking app)* Renovate AI (home design tool)* Post AI (Washington Post’s chatbot experiment)* Notebook LM by Google* Huxe (new personalized podcast app from former Notebook LM team)* Ollama (run local LLMs)* Reins (GUI app for local model management) This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit productfridays.substack.com
Transcribe →