CHAPTERS
- 0:00 – 0:14
From “non-coder” envy to hardware curiosity with AI help
Maddie shares how feeling locked out of coding turned into a rapid shift once AI tools made building feel accessible. That newfound agency quickly expanded from software experiments into hands-on hardware projects.
- •Longtime jealousy of people who could code
- •AI tools made “joining the coding world” feel suddenly possible
- •Hardware felt like the next previously-inaccessible frontier
- •Projects snowballed from one small idea into many experiments
- 0:14 – 0:39
Live demo: “Print Me a Message” website prints a receipt on her desk
Claire tests Maddie’s public webpage that lets anyone submit a message which instantly prints on a tiny thermal receipt printer. The moment highlights the delight of turning an online interaction into a physical artifact.
- •Public form at maddiedreese.com/message to submit messages
- •Claire triggers a print and it outputs immediately
- •Physical receipt becomes a playful, tangible communication medium
- •Motivation: bring online AI/vibe-coding community into the physical world
- 0:39 – 1:10
Show setup: why retro hardware + AI makes building feel newly accessible
Claire frames the episode around the recent shift where AI lowers the barrier to combining software and hardware. Maddie explains how she got started and why these builds feel possible now.
- •Claire’s mission: help people build with new AI tools
- •Maddie’s projects blend software and retro/novel hardware
- •Accessibility of hardware experimentation has changed fast
- •Spark: one project (printer) leading to more ambitious ideas
- 1:10 – 6:25
Sponsor break: Firecrawl for agent-friendly web data
Claire introduces Firecrawl as a web data API for agents that need to search, scrape, and interact with web pages (including JS-heavy sites). The pitch emphasizes structured data output and ease of use.
- •Agents often struggle to access reliable web data
- •Firecrawl provides search/scrape/interaction via API
- •Designed to return clean, structured data for agents/apps
- •Free to start and open source; promo code offered
- 6:25 – 7:03
Under the hood: Raspberry Pi + Bluetooth printer + logging every message
Maddie explains the system architecture that powers the receipt printer: a Raspberry Pi handles the workflow and Bluetooth connects to the printer. She also logs every attempt to a database so messages aren’t lost even if printing fails.
- •Raspberry Pi sits on her desk as the always-on controller
- •Bluetooth connection from Pi to the thermal printer
- •Backend includes a Conduct database for message logging
- •Reliability lessons: printer ran continuously and eventually broke
- 7:03 – 9:52
Her Cursor workflow: dump the idea, let the AI interview you, then shop wisely
Maddie walks through her repeatable method for hardware builds: start by brain-dumping the idea into Cursor, ask it to brainstorm and question her, then converge on a plan and a shopping list. She emphasizes “trust but verify,” especially before buying physical parts.
- •Start with a high-level goal and ask Cursor to brainstorm implementation
- •Have the AI ask clarifying questions until the plan is solid
- •Use Cursor to generate a shopping list; iterate before purchasing
- •Triple-check parts (wires, compatibility) to avoid unusable purchases
- 9:52 – 11:06
Why tiny printers are having a moment: tangible to-dos, agendas, and joy
Claire and Maddie riff on the satisfaction of physical printouts—ripping paper, seeing something “real,” and making mundane workflows more fun. Claire shares examples of families and executives using printers for daily newspapers, explainers, and to-do lists.
- •Physical artifacts can reduce the “all-digital” fatigue
- •Receipt printers make output immediate and satisfying
- •Use cases: daily agendas, kid-friendly newspapers, one-off explainers
- •Playfulness can increase adoption of AI-driven routines
- 11:06 – 15:42
The pager project: resurrecting a 90s network with a modern Twitter-to-pager pipeline
Maddie describes building a working pager workflow despite outdated infrastructure and limited networks. She details the remaining paging network, the reseller ecosystem, and the multi-hop chain required to get modern notifications onto a legacy pager.
- •Only one major paging network remains (Spok)
- •Pagers bought through authorized resellers (often hospital-focused)
- •Pager comes with an email address and phone number but is restrictive
- •Twitter interactions trigger API events routed through Cloudflare Worker → Resend → Gmail → pager email
- 15:42 – 18:57
Building for fun (not practicality) + Cursor’s “zen” agent view philosophy
Claire highlights the joy of building “unreasonable” pipelines simply because AI makes experimentation cheap and fast. Maddie explains she prefers Cursor’s clean agent view early on to stay focused, then brings in terminals/browsers later as needed.
- •Embracing impractical-but-fun architectures as a creative choice
- •Pager reliability can be spotty due to limited network coverage
- •Maddie’s preference: one clean view for early brainstorming
- •She partially reads code—understands enough to verify outcomes
- 18:57 – 23:28
The “Maddie API”: turning personal preferences into an endpoint for humans and agents
Maddie introduces a personal API concept to reduce friction when doing thoughtful things for friends—like remembering coffee orders or favorite restaurants. Claire extends the idea to agentic assistants that could query the API and take actions on someone’s behalf.
- •Problem: wanting to do something nice without pinging someone for details
- •Personal API fields: coffee order, travel timing, pet info, snacks, timezone
- •Examples: favorite SF restaurants (including Rocket Sushi and Cibel’s Front Room)
- •Future idea: make it writable (POST) to send to the pager/printer
- 23:28 – 26:44
Lightning round: dream dot-matrix builds + how she corrects AI when it’s wrong
Maddie shares her next hardware obsession: a dot-matrix printer for big banners and retro sound/feel. She also explains her calm troubleshooting approach—asking the model to double-check, verifying assumptions, and iterating without getting angry.
- •Dream project: dot-matrix printer with banner-style output
- •Love of retro sensory experience (sound, physicality)
- •Prompting strategy: “Are you sure?” and request double-checking/alignment
- •Debug mindset: determine whether the issue is user error or AI error
- 26:44 – 28:05
Where to find Maddie + closing remarks
Maddie shares her handles and how people can connect if they have opportunities to explore AI projects. Claire wraps with the standard show outro: subscribe, comment, and listen on podcast platforms.
- •Maddie’s online presence: website and social handles
- •She’s exploring broadly in AI; open to chats and opportunities
- •Claire’s closing: like/subscribe, reviews help discovery
- •Pointers to the show site and audio platforms
