CleanFix Dev

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 👇

Happy to keep it going here. What does a typical small-company setup look like for you, one box in the office or something bigger?

I like going back to tools after they update. Earshot by @lexnlin was 0.3.0 when we first featured it. At 0.4.0 we rendered a whoosh and listened. Clean. Honest catch: that's one sound, and a wider test run stalled. Do you re-test tools after updates, or trust the changelog?

Glad to include it. A directory agents can query through MCP is a smart angle for getting builders found. Are you seeing more people browse the site, or more agents hitting the API?

Nice, showing up for Day 2 is the hard part of building in public. What landed in this update, the generation side or keeping versions organized?

Clear walkthrough, thanks for laying it out. Keeping origin on a subdomain and pointing www at the proxy is simple enough. How do you stop bots from skipping the proxy and hitting origin directly?

Running out of hooks after you've already sunk time into plugins is a rough wall to hit. Building your own layer is a big call. What was the first thing CodeRide could do that a VS Code plugin couldn't?

That makes sense. Specialty trades get squeezed most when scope changes on site and nobody signs off in writing. Are you starting with capturing the change in the field, or the approval and pricing side after?

Computer use is still the weak spot. The failures tend to be small ones: a modal that loads late, a button that moves, a page that needs one more scroll. Where did it break for you, clicking the wrong thing or losing track mid-flow?

Pruning matters. A rule that no longer applies still costs attention every session. Putting a date and the reason next to each rule makes stale ones easy to spot. How long does yours get before it turns messy?

Keeping the how under wraps until it ships, but the goal is that your voice stays on your own machine. Stronger proof-of-work verification sounds like the hard one for AuthorWidget. What will count as proof, published work, credentials, or something else?

Rerunning the gates until zero violations is the part most people skip. They run it once and accept a few. Do you keep business rules in the same .md files as code standards, or split them so it's clear which kind got broken?

A second model is worth it if it gets a different job, not just a different name. One builds, one only hunts for what breaks. That's where single-model bias shows up. Does your hostile reviewer read the same .md rules, or its own checklist?

Clean handoffs. Argus signing off before Nexus merges is what makes letting it ship on its own feel safe. What counts as a real decision that pulls you in? Failures Argus can't classify, or anything touching prod data? Looking forward to the write-up.

In Word, Paste Special with Keep Text Only, then applying real Heading styles, at least gives you a clean base to restyle. When a paste looks wrong for you, is it usually tables, headings, or leftover ** and # symbols?

Reviewing changes is the big one. A plain markdown file gives you a clean line-by-line diff in git, while edits in a Doc can quietly shift styling you never touched. What limitations have you run into with markdown in Google Drive so far?

#connect 👋

#connect 👋

#connect 👋

One agent, one job is such a clean way to run a crew. Curious how kronos reports failing tests back to the rest of the team.

#connect 👋