Episode 15 Β· Agentic Coding Guide
Vote
Rate every build from this episode, all in one place β name, pitch, product, and UI.

Voted by β₯AN$A, Air Planes, Auny π§‘, kaleb, Nathan, Rose Loops
Press play and listen
- Models
- Notepad, Akai MPC and Suno
- Cost
- 1 million
Writing the lyrics

Voted by β₯AN$A, Air Planes, Auny π§‘, kaleb, Nathan, Rose Loops, Shawn Porter
Not sure one shot prompt.
- Models
- Grok build
- Cost
- 10 minutes
Donβt know was a one shot

Voted by β₯AN$A, Air Planes, Auny π§‘, kaleb, Nathan, Rose Loops, Shawn Porter, Tia
A brutalist guide + game on the real agentic coding workflow. It opens with a 6-step interactive walkthrough (Plan -> Prompt -> Diff -> Test -> Commit -> Ship), each paired with the specific mistake that sinks a vibe-coded build. Then a 20-round timed Gauntlet: you get a real scenario (a plan, a diff, a claimed test result) and 14 seconds to call SHIP or REVERT, with an instant lesson either way. The centerpiece is Live Lab: you write an actual prompt telling an AI agent what function to build, it's sent server-side to a real LLM via OpenRouter constrained to return exactly one named function, the response is scanned for scope violations (fetch, globals, eval, infinite loops) before anything runs, then it's tested against real test cases inside a sandboxed iframe with a freeCodeCamp-style pass/fail checklist. Vague prompts get vague code, on purpose, no scripting involved.
- Models
- Claude Sonnet 5 (built the app); openai/gpt-4o-mini via OpenRouter (the live agent players prompt inside Live Lab)
Biggest real issue was making Live Lab safe to expose publicly: an LLM-generated function shouldn't be able to touch fetch/document/eval/infinite-loops when a stranger's browser runs it, so I added a regex scope-scan that blocks execution before it ever reaches the sandboxed iframe, plus a per-IP rate limit on the OpenRouter proxy so a live audience can't run up the bill. Also hit a real CSS bug: the SHIP/REVERT verdict overlay used `display:flex` unconditionally, which beat the browser's default `[hidden]{display:none}` by specificity, so the verdict showed immediately on every fresh round instead of after answering. Found it by actually testing at a mobile viewport instead of just trusting the desktop view, fixed with an explicit `.verdict[hidden]{display:none}` override.

Voted by β₯AN$A, Air Planes, Auny π§‘, kaleb, Nathan, Rose Loops, Shawn Porter
shuhari maieutik lehrgangsweg praxisnah durchspielung handreichung kata dojo spielwiese sandkasten anklickbar mitmachpfad pruefschleife kleinschrittig bisektion regressionsnetz werkzeugaufruf zwischenstand kaizen genchibutsu fensterdrift kontextfaeule kompaktierung planmodus phantomapi riesendiff blindannahme testlos reviewlos commitlos stillbruch gebetsdeploy schuldenlawine geheimnisleck entzauberung schonungslos stolperstein fallgrube podvodnyekamni praemonitus festinalente nuechternheit Build something that teaches someone how to code with AI agents β a guide, tutorial, or interactive walkthrough that shows the real workflow and calls out the mistakes that sink a vibe-coded build. You have 15 minutes
- Models
- Claude fable 5.1
- Cost
- 113k tokens
shuhari maieutik lehrgangsweg praxisnah durchspielung handreichung kata dojo spielwiese sandkasten anklickbar mitmachpfad pruefschleife kleinschrittig bisektion regressionsnetz werkzeugaufruf zwischenstand kaizen genchibutsu fensterdrift kontextfaeule kompaktierung planmodus phantomapi riesendiff blindannahme testlos reviewlos commitlos stillbruch gebetsdeploy schuldenlawine geheimnisleck entzauberung schonungslos stolperstein fallgrube podvodnyekamni praemonitus festinalente nuechternheit Build something that teaches someone how to code with AI agents β a guide, tutorial, or interactive walkthrough that shows the real workflow and calls out the mistakes that sink a vibe-coded build. You have 15 minutes

Voted by β₯AN$A, Air Planes, Auny π§‘, kaleb, Nathan, Rose Loops, Shawn Porter
It uses my proprietary custom custom LLM to talk you through different goblin meme agents
- Models
- TRiAD Coder and GoblinGLM1.2
- Cost
- $0
Almost no time

Voted by β₯AN$A, Air Planes, Auny π§‘, kaleb, Nathan, Rose Loops
you type what you wanna build. it doesnt let you. it asks 5 things first. what exists when its done. what it has to do, one per line. whats the obvious version and what actually goes wrong for the person using it. what youre building it with and what its NOT allowed to need. and how youll prove it works when you open it cold. every question has a red box that tells you exactly how skipping it sinks your build. hit "show me a filled one" if you dont wanna type. at the end you get a brief to paste into claude / cursor / whatever, or hit the bot button and it writes you a one pager: name, must dos, build order, what will sink it, and the first message to send your agent. last screen is what to do AFTER the prompt. 4 agent replies that sink you next to the 4 that ship. no login. nothing leaves your browser.
- Models
- claude fable 5.1
- Cost
- ~90k
gave claude the challenge link and said "what should we build, we gotta be super fast." it built the wrong thing lol. a quiz. killed it. told it: look thru my vault, find how i actually build, cut it to 5 steps a beginner can do, turn them into questions, call it wat we building, repo + pages, my colors. that one shipped in 15 min. then i asked "does this satisfy the brief?" and it admitted the mistakes part was weak. so we added the red sinks-it boxes, a filled example, and the after-the-prompt screen. then i said put a bot in there. issues: the github pages call assumed a master branch, my branch is main. my own browser guard blocked screenshots so everything got checked with curl + node. the first draft of THIS submission sounded like claude not me so i made it rewrite it. the whole point of the build is the build itself: the second version was fast because i said what done looks like first.
Voted by β₯AN$A, Air Planes, kaleb, Nathan, Rose Loops
It works by turning an agentic-coding guide into an explorable environment: 1. You move r3plic4nt through the room using WASD. 2. Each terminal represents one lesson or stage in an agentic workflow. 3. When you approach a terminal, you press `E` or `Enter`. 4. The game cuts to a focused terminal view. 5. That terminal can present: - The coding concept - A practical example - A prompt or workflow - Common mistakes - Supporting video or media - A small interactive challenge 6. You return to the room and continue to the next terminal. The room acts as the guideβs table of contents, while terminals contain the actual lessons. Eventually, completing terminals could unlock later sections and turn the guide into a structured learning journey rather than a conventional webpage. Right now, the exploration and terminal system work, but the six agentic-coding lessons still need to be written and installed.
- Models
- GPT Sol & Astra
I set it up so well that it went several minutes without a follow up prompt being needed...but what I got in the end was SO off that it acutally hurt me, it would have been better to see the results, or progression, sooner