AI and Coding with ChatGPT

Episode Summary

Executive Summary: The episode examines the rapidly evolving AI landscape for developers and designers, focusing on tools like ChatGPT, GitHub Copilot, Midjourney, and Bing’s AI search. Wes and Scott argue AI won’t simply replace jobs, but it will change workflows, raise the baseline for quality, and make prompt-writing, judgment, and creativity increasingly important. They also discuss model training, ethical concerns, accessibility, and how developers should start integrating AI APIs into applications.

Main Topics: AI tools available to developers and creators: They survey the current ecosystem: ChatGPT for conversational assistance, GitHub Copilot for code completion, Midjourney and Stable Diffusion for image generation, Bing for live web-backed answers, and TensorFlow/OpenAI APIs for custom model work. How AI changes coding workflows: The hosts describe practical use cases such as generating dummy data, regexes, TypeScript types, query parsing examples, and code refactors, emphasizing AI as a productivity booster rather than a full replacement. AI’s impact on jobs and skill value: They reject extreme takes that AI will either destroy everything or do nothing. Instead, they argue it will change jobs, increase automation, and make creativity, problem-solving, and prompting more valuable. Prompting as a new core skill: They frame prompting as a form of coding and information retrieval, where asking the right question becomes essential to getting useful output from AI systems. Ethics, attribution, and bias in AI-generated content: They discuss concerns about training data being scraped without attribution, AI-generated website and art styles resembling existing work, and the risks of misinformation and synthetic media. Specialized models and on-device machine learning: They note that ML is already embedded in everyday products like autocorrect, hearing aids, and smart-home devices, suggesting AI will spread beyond web apps into physical tools and appliances. Design quality, accessibility, and the raising of the baseline: They argue tools like Canva and AI-assisted design are making average output better, while custom branding, accessibility, and human touch become differentiators.

Key Arguments: AI is not a fad; it is becoming a durable part of development workflows and broader digital work. The most realistic near-term effect is augmentation: faster coding, better scaffolding, and less time spent on repetitive tasks. Developers must still understand the code AI produces, because blindly copying it can introduce bugs or harmful behavior. Prompt quality matters: knowing how to ask, refine, and constrain AI is becoming a practical professional skill. AI increases the baseline quality of design and content, which makes originality, taste, and custom human work more valuable. There are serious ethical and trust issues around training data, attribution, and AI-generated misinformation. AI features will increasingly be expected in applications, so developers need to learn how to integrate AI via APIs and models. Lower-cost tools and subscriptions may widen access gaps, since many powerful AI services are paywalled.

Data Points: ChatGPT training cutoff: up until 2021 - Described as the knowledge boundary of ChatGPT during discussion of its capabilities. ChatGPT model size: 175 billion parameters - Mentioned as the scale of the GPT-3/ChatGPT model. Model data size: 45 terabytes - Referenced via Google as an estimate of the underlying data scale. Midjourney subscription cost: $10/month - Scott mentions paying monthly for Midjourney. Copilot cost: $8-$10 - Referenced as the typical monthly price for GitHub Copilot. Sentry promo: 2 months free - Coupon code TastyTreat offered for the sponsor. AI-generated app size: 2.5 GB - Amazing AI desktop app bundled with Stable Diffusion was described as roughly this size. Office motion sensor price: $24 - Scott’s Aqara motion sensor for smart-home automation. Snowboarding top speed: 41 miles per hour - Scott’s watch/app tracked this speed during snowboarding. Snowboarding distance: 14 miles - Scott reported total distance covered during the session. Dummy data generation: 50 people with nested jobs and date ranges - Example prompt Scott uses with Copilot to generate realistic course data.

Pivotal Quotes: "It's not going to go away." — Wes: Used to argue AI is durable and should not be dismissed as a fad. "It's literally a booster pack. It's a rocket pack to coding." — Scott: Describing how GitHub Copilot accelerates coding productivity. "Prompting and how to get the information or the result you want out is going to be a big deal." — Scott: Explaining why prompt-writing will become an important technical and creative skill.

Implications: Listeners should expect AI to become standard in development, design, and everyday tools. Success will depend less on rote production and more on judgment, prompting, creativity, and the ability to verify AI output.