This transcript captures a series of conversations from the #share-it channel of the "Solve It With Code" Discord community. The discussions center on members' experiences using SolveIt, an AI-powered coding platform, to tackle a wide range of projects. Users share their work, which includes solving Advent of Code challenges, building a Lisp interpreter from scratch, developing a recursive descent parser, and creating complex infrastructure setups involving SSH, Tailscale, and Cloudflare Tunnels within the SolveIt environment.
A recurring theme is the platform's impact on the learning and coding process. Members highlight the effectiveness of its deliberate, step-by-step methodology, the ability to get "unstuck" from difficult problems, and the power of directly editing AI-generated output. The transcript also showcases community engagement through discussions on advanced features like "tool calling," different operational modes, and questions about scaling applications for production use. One member even shares a project where they built a minimal clone of the SolveIt platform to better understand its architecture, receiving praise from the community.
Highlights from the Solve It With Code Community Showcase
Nov 5, 10:45 PM Key Technique: Direct Editing: A user shares their success in refining an article by directly editing the AI's output in SolveIt, rather than trying to correct it through further prompting. Instructor Jeremy Howard reinforces this as a best practice, highlighting it as a more efficient workflow.
Nov 6, 12:05 AM Learning Through Deliberate Practice: A user details their "breakthrough" moment in understanding recursive descent parsers while building one. They emphasize that the platform's line-by-line, deliberate practice approach helped solidify complex concepts that were previously abstract.
Nov 6, 2:21 AM Exploring Agentic Capabilities: A member demonstrates the "tool calling" feature by having the AI execute a custom Python function multiple times without having access to its source code. This showcases SolveIt's ability to act as an agent that can use developer-defined tools.
Nov 6, 6:54 AM An Enhanced Learning Workflow: After building a Lisp interpreter, a user lists several key benefits of the SolveIt workflow: it helps organize steps, maintains flow by handling minor syntax questions, guides good project trade-offs, and provides perfectly-timed hints to overcome frustration.
Nov 6, 10:02 AM Planning for Production: A community member's question about scaling applications leads to a discussion on deployment strategies. The advice given is to use Answer.AI's platform (pls.sh) for small to medium scale and dedicated servers for high-scale production, indicating that users are building real-world applications.
Nov 6, 8:55 PM Rebuilding SolveIt from Scratch: A user shares a highly praised project where they created a minimal, 200-line clone of the SolveIt platform. The goal was to understand its core technologies, including its HTMX-powered UI, persistent Python interpreter, and context-aware LLM integration.
Nov 7, 5:06 AM Using Different AI Models: A user shares their journey of building a sentiment analysis tool, starting with no knowledge. When they couldn't use the default model (Claude) due to its free tier limitations, they used SolveIt to guide them in successfully integrating Google's Gemini API instead.
Nov 7, 7:50 AM Advanced Infrastructure Integration: A user presents a sophisticated project that runs SSH, Tailscale (for secure networking), WebDav (for file sharing), and Cloudflare Tunnels directly within the SolveIt environment. The ultimate goal is to use SolveIt as a development environment to interactively run code on other compute resources.