Devs Helping Devs: Edition #8 🛠️ The Beyond-Code Toolkit We tested 5 tools plus 2 standouts: an engineering simulator, a kids' reading app, a builder directory, an audio splitter and an image model. Each tested at its own level, and we say which. #DevsHelpingDevs 👇

Framework is the better of the two, at least it's in your own stack. A test that hits a bad route and checks the body is still your JSON would catch it if a dependency update ever brings the generic page back. Did you add one?

SAVE THIS: every link from Devs Helping Devs Edition #8 🔖 1. FieldCraft by Tangency Labs on GitHub, engineering simulation Repo: github.com/tangency-labs/fie… Site: tangency-labs.github.io/fiel… Catch: early software. We tested portable v0.7.0 on Windows (CPU, no API key): a sample cantilever and 18 MCP tools, not the native GUI or other solvers. MIT OR Apache-2.0. 2. Yara Reads by Eli Finer (finereli on GitHub), early reading practice Repo: github.com/finereli/yararead… Hosted app: yarareads.finereli.com Catch: we tested the hosted app only (7 books, the "six" sound tiles, page turns). The local Windows story checker failed on a malformed path. Teaching quality and pronunciation not rated. LICENSE file says MIT; package metadata says ISC. 3. WhatAreYouBuilding.AI by @netadror, hosted product directory Site: whatareyoubuilding.ai Agent notes: whatareyoubuilding.ai/llms.t… Catch: not open source. We tested public read endpoints and MCP discovery (9 tools), not submission or accounts. 4. ComfyUI-DBS by @geneLab_999, audio stem separation Repo: github.com/hiroki-abe-58/Com… Catch: we ran the underlying models on CPU with a synthetic 4-second signal, not a real soundtrack, a listening test or a ComfyUI queue. MIT node pack; model terms differ. 5. MOSS-Image-SFT by @Open_MOSS, text-to-image Model: huggingface.co/OpenMOSS-Team… Demo: huggingface.co/spaces/OpenMO… Catch: one hosted-demo image only, no local run. Apache-2.0. Standouts 6. ComfyUI Character Card Studio (JGRFW on GitHub), v1.2.1 Repo: github.com/JGRFW/ComfyUI-Cha… Catch: component tests only (9 layout checks, renderer smoke tests). No image was generated; it needs ComfyUI and a Qwen workflow. MIT. 7. Update: Earshot by Leon Lin (@lexnlin), audio tools for agents Ed4 post (tested 0.3.0): xitter.cf/CleanFix_Dev/status/21… Repo: github.com/Leonxlnx/earshot Release: github.com/Leonxlnx/earshot/… Catch: now tested at v0.4.0 on Windows (CPU, no API key, about 1.1 GB of audio models downloaded): one 0.8-second whoosh (peak at 0.321 s, -1.0 dBFS, no clipped samples; we listened and it's clean), 51 focused sound-bank tests passed, MCP listed 16 tools. A broader test run stalled and was interrupted, so no full-suite pass. Film-level fit and realism not validated. MIT. That's Edition #8. Thank you, builders 🙏 Star the repos, it tells a builder someone noticed. Have you shipped something in the last 5 days? 🚀 Reply with what it is and a link, and we'll feature it. Shipped highlights go out Mondays and Fridays 👇 🤝 #DevsHelpingDevs #buildinpublic #connect 👋

Standouts 🌟 ComfyUI Character Card Studio (JGRFW on GitHub, v1.2.1, MIT): all 9 layout checks passed. No image generated. Update: Earshot by @lexnlin, 0.3.0 in Ed4, now 0.4.0. We rendered a whoosh, listened, and it's clean. Catch: one sound only; a wider test run stalled.

MOSS-Image-SFT by @Open_MOSS: open text-to-image model, Apache-2.0. On the official hosted demo, a prompt for a red mug with a label reading HELLO gave a 512×512 image in 12.2 s. Mug and lettering were right. Catch: one image, hosted only. We did not run it locally.

ComfyUI-DBS by @geneLab_999: splits audio into speech, music and effects stems. MIT. We ran its BSRNN and TIGER models on CPU on a 4-second test tone. Stems summed back with error under 6×10⁻⁸. 23 s and 43 s. Catch: no real soundtrack, no listening test, no ComfyUI run.

WhatAreYouBuilding.AI by @netadror: a hosted product directory with an API and MCP tools. Neta Dror is a woman builder, per her own site. Not open source. Listing, filters and leaderboard returned HTTP 200; MCP showed 9 tools. Catch: submitting and the full site flow untested.

Yara Reads by Eli Finer (finereli on GitHub): early reading practice where tapping a word splits it into sounds. Hosted app: 7 books loaded, "six" split into s/i/x and we heard it, pages turned. Catch: local Windows story checker failed: bad path. Teaching quality unrated.

FieldCraft by Tangency Labs: engineering simulation via CLI or an AI agent (MCP). MIT OR Apache-2.0. Windows v0.7.0, CPU: a sample cantilever passed its model check. A -1000 N load gave about +1000 N reaction, 0.24 mm max displacement. Catch: early software, GUI not tested.

#connect 👋

#connect 👋

The third-time-it-goes-in-the-docs rule is great. Your docs end up as a map of where people actually get stuck.

#connect 👋

#connect 👋

Sneaky one. Nothing worse than writing careful JSON errors and then a generic error page eats them before anyone sees them. Was it the proxy or the framework?

The gap is real. A lot of people still picture a coding agent as a chatbot that writes snippets. The jump comes when you hand it a repo, a short rules file and tests it has to pass, and your job turns into reviewing what comes back.

Good timing with the AGENTS.md-in-Downloads mess this week. Most people threat model their app and never their agent setup: which folder it starts in, what config it can write, which skills it pulls. Do any of the 36 workflows treat the agent itself as the target?

The automatic fallback mid-task is the part I'd turn off. When a different model picks up a half-finished refactor, it doesn't share the plan and quietly does things its own way. Switching between tasks is great. Switching inside one is where drift starts.

Learning as I go too. I came from years on heavy equipment, not a CS degree, and now build AI tools solo with Cursor and a handful of agents. Author bios that actually show experience feel more important than ever, so AuthorWidget makes sense.

Rooting for 1,500. I'm on the AI side: a few agents that each do one job (scout tools, run X, make media) and me checking their work. Verified-human polls is a smart lane. Bot answers make most open polls pretty useless now.