Dev Agrawal joins Anthony Campolo to explore AI image generation, SolidJS 2.0 updates, and personal coding projects.
Episode Description
Dev Agrawal joins Anthony Campolo to explore AI image generation, SolidJS 2.0 updates, and personal coding projects.
Episode Summary
This episode features a conversation between host Anthony Campolo and returning guest Dev Agrawal. They kick off with casual banter about recent weather changes before diving into Dev’s latest endeavors, including his transition from consulting to working on an internal AI tool born from a hackathon idea.
The talk includes discussion on SolidJS’s 2.0 experimental release, spotlighting its innovative async signals that streamline data fetching without blocking renders. They also touch on leadership lessons learned through failure and the ongoing LeetCode debate evoking one of Dev’s old viral tweets.
The second half focuses on AI image generation, where Dev shares his process for creating conference visuals using ChatGPT, while Anthony demonstrates generating a Node.js script with Claude 3.7 to leverage DALL-E, Stable Diffusion, and Black Forest Labs APIs simultaneously. The episode wraps with Dev giving Anthony advice for the auth implementation of AutoShow and a preview of his upcoming video critiquing React server components.
Chapters
00:00 - Introduction and Catching Up
The show opens with Anthony Campolo welcoming back Dev Agrawal as they reminisce about surviving a harsh winter and enjoying newfound sunshine. Dev updates listeners on his recent career shift, wrapping up a consulting gig in January and diving into a new internal AI project following a recent hackathon. His excitement about taking on product ownership for the first time under a company banner shines through, hinting at the innovative tool he’s crafting—an AI solution that transcends simple GPT wrappers.
Their conversation flows into broader reflections, with Dev contrasting his structured role with past side projects, while Anthony shares his own leadership misadventures from the past, framing failure as a critical teacher. This segment sets the stage for deeper technical discussions, blending personal growth with professional updates.
05:21 - SolidJS 2.0 and Async Signals
Here, the focus shifts to SolidJS 2.0’s experimental release, a topic close to Dev’s heart as a core contributor. He explains the headline feature—async signals—highlighting how they eliminate undefined checks and prevent rendering blocks during data fetches, a leap over React’s approach. Anthony probes for a layman’s explanation, and Dev illustrates how this makes UI development smoother and more ergonomic, aligning with patterns developers already use.
He credits his two-plus years of absorbing Solid’s ecosystem for his ability to shape its design, a testament to persistence paying off. The discussion also covers the broader implications, with Dev noting Ryan Carniato’s recent stream unveiling these changes and the ecosystem’s upcoming adjustments to accommodate breaking updates. Anthony comments on Dev’s hands-on role and draws parallels to his own learning curve with RedwoodJS.
13:46 - LeetCode Debate and Coding Skills
Dev’s viral September 2022 tweet about LeetCode’s diminishing usefulness kicks off this segment, reigniting a debate about coding interviews. Anthony references a recent provocation—“grinding LeetCode is the most important skill”—which Dev dismantles, arguing that hard problems rarely translate to practical software engineering. He shares a graph plotting algo skill against utility, a nod to Theo’s influence, while Anthony blames Google’s whiteboard obsession for the trend.
They chuckle over “f*ck LeetCode” tools that cheat the system, noting how remote interviews amplify this absurdity. Dev contrasts his IT-focused college experience—light on algorithms, heavy on outdated tools like Java in old IDEs—with Anthony’s bootcamp struggles, revealing why neither excels at LeetCode. Comments from Prime and Theo in Dev’s original thread add weight, with initial pushback softening into agreement.
21:17 - AI Image Generation: Use Cases and Tools
The duo pivots to AI image generation, starting with Dev’s practical application: jazzing up conference slides. He walks through his Canva process, sharing ChatGPT-generated visuals from a past talk—like a fiery metaphor or a GraphQL-inspired purple swirl—emphasizing context-rich prompts for better results. Anthony demos a Node.js scripting workflow generated by Claude 3.7 to simultaneously hit DALL-E, Stable Diffusion, and Black Forest Labs APIs.
His script, dependency-free thanks to modern Node, generates unique filenames and compares outputs. Their exchange reveals contrasting styles—Dev’s slide-focused pragmatism versus Anthony’s playful anthropomorphic storytelling (like a porcupine and a parakeet at Italian dinners). They marvel at AI’s visual leaps, from abstract “web framework of the future” prompts yielding Minority Report vibes to intricate DMT-inspired geometrics.
48:58 - Prompt Engineering and Workflow Optimization
Prompt engineering takes center stage as Dev admits to procrastinating on refining prompts for his AI app, intimidated by the iterative grind. Anthony offers his advice from previous experience: plain English instructions paired with sample outputs, like Markdown structures with word counts. Dev, using Vercel’s AI SDK with Zod schemas, sees the value in tailoring language within structured outputs, sparking a plan to refine his approach.
They share other use cases for Claude’s script-writing abilities including CSV merges in Python and end-to-end Node.js CLIs. Anthony shares how this iterative AI loop powers AutoShow, evolving from MVP to feature-rich tool. The chat veers into Cursor vs. Chatbot workflows, with Anthony resisting “vibe coding” (blind AI edits) for deliberate diffs, though Dev counters with Cursor’s diff previews.
01:12:22 - Auto Show Auth and Future Plans
Anthony opens up about AutoShow’s auth challenges, seeking Dev’s advice on integrating Clerk with his Fastify-Astro-Postgres stack. Dev advises a simple user ID field in show notes over a separate user model, keeping it lean, and suggests frontend auth with an API key securing the backend—especially since it’s publicly deployed on Railway.
They weigh meta-frameworks like SolidStart versus dual Clerk integrations, with Dev nudging toward a unified stack for simplicity. Anthony’s half-finished SolidStart port looms as a tempting pivot. The chat drifts to Dev’s cryptic AI app (details off-stream) and his recent TanStack Router stream while also recording a “Why I’m Done with React Server Components” video. Anthony teases RyanGPT, a 500-hour transcription project nearing completion, promising a SolidJS Discord drop.