March 21, 2026
Bots, bytes, and brawls
Show HN: AI SDLC Scaffold, repo template for AI-assisted software development
AI starter kit drops: skeptics want receipts, fans want Figma
TLDR: An AI-led project starter kit promises to let bots handle the build while humans supervise, with every decision logged in the repo. The crowd split fast: skeptics want hard numbers to prove it won’t waste money, while fans cheer the concept and ask for Figma integration to bridge design and code.
A new “AI-first” project starter kit just hit Hacker News, promising to let bots do the heavy lifting while humans steer. The repo template lays out four clear stages—Objectives, Design, Code, Deploy—and keeps everything in one place so AI tools can follow instructions, make choices, and log every decision for review. Translation: you manage the big picture, the AI handles the grind. It even offers a quick start via degit and strongly warns: no manual edits—let the agent do it.
Cue the comments. One camp came in blazing with spreadsheets and suspicion. User zihotki demanded benchmarks and evaluations, basically asking, “Does this save money or just burn tokens?” In plain speak: if the AI keeps chatting, someone’s credit card is crying. The other vibe was the design crowd dreaming bigger. gzoo cheered the idea and wished for a Figma hookup—because if the AI is building, why not let it read your designs too? Figma fans, assemble.
So the thread split into two classic internet tribes: The Receipts Crew (“Prove it!”) versus The Optimists (“Ship it!”). Jokes rolled in about “AI interns” pushing commits at 3 a.m. and “token bonfires” if prompts get wordy. Love it or side-eye it, the crowd made one thing clear: AI doing the busywork is hot—but only if it’s efficient and plays nice with design tools.
Key Points
- •AI SDLC Scaffold is a repository template for AI-first development with Claude Code, organizing work into Objectives, Design, Code, and Deploy phases.
- •The scaffold’s principles include AI-first workflow, everything-in-repo, context-window efficiency, and comprehensive decision capture.
- •Quick start recommends using degit (requires Node.js) to copy the scaffold without Git history; manual copy steps are also provided.
- •Users run /SDLC-init in Claude Code to set up project context, then proceed through phases using built-in skills or custom prompts.
- •Manual edits are discouraged; all changes should go through AI agents following repository instructions to ensure consistency and recorded decisions.