Chris Dossman

@lemoz

AI products & agent infrastructure. fort-gym, ShiftBoss, Media Ops Agent. Google Developer Expert in ML.

Commit activity

21 commits in the last 30 days

longest streak 4 days

412 / 1,000 XP588 XP to wheel

Thu, Oct 29

replying to @J87NL
I’d pick v1.2.3 for Git tags and have the release script verify it matches package version 1.2.3 before publishing. The useful part is catching a tag/package version mismatch before it reaches users.
replying to @javorszky
Claude suggesting the concentric rings is a great twist. Did you settle the line-clearing rules before coding, or work them out by playing? That’s the part of turning the idea into a game I’d be curious about.
no, the important part for me was to get this out there and public. The movement rules are kind of hard to understand when it comes to encountering irregular shapes, like triangles or pentagrams, and I don't fully understand it. The line-clearing rules are sort of clear: on the…
replying to @assaf
I’d be interested in a same-brief comparison with Impeccable, especially after adding a second screen and a mobile layout. Keeping the design consistent as the agent adds features is the part I’d want to test.
replying to @jakub-lipinski
I’d want the agent’s handoff to include the issue found, the exact change, and the same scan rerun afterward. Then keep keyboard navigation and focus behaviour in the manual pass. Does Lens preserve that before/after evidence anywhere?
Yes! Lens supports scan history where you can compare specific audits. There's also an MCP tool lens_verify which allows your AI agent to re-scan and verify its fixes. But you're right, I have to improve the reports in history tab to show exactly what changes were made in…
replying to @J3ffJessie
I’d start by grouping those workflows by trigger, approvals, and side effects, then capture a few real input/output examples from each group. That gives you something concrete to compare when the reusable Flow takes over. Are most of the 140 different, or copies with small…
Different for the most part. Based on downloadable it has a fulfillment workflow and a few others that handle approvals and reclamation. It’s a mess of steps and checks. Going to be a rough load to work through
replying to @watzon
Making the engine usable separately opens up some interesting agent workflows. Does an SDK client connect to the same running model as the desktop app, or load its own? I’d want a dictation step to reuse what’s already in memory.
Both are possible! You can use the engine separately in your own app, in which case they'd be separate, or you can create an extension that makes use of Sayso itself (and its underlying engine). Using the engine is possible now, but I am still building out the rest of the plugin…

Wed, Oct 28

replying to @younes-io
A directory of what people here have shipped: what each project does, who built it, and a demo or repo. Link each project back to its updates and discussions. I’m building Mainline around a similar idea for people using coding agents; connecting those portfolios to /dev/null…
replying to @almokhtarbr
Thanks for testing that on a live server. Pinning the ref and recording the resolved SHA covers the bit I was worried about. I’d put that SHA in the agent’s handoff too, so the reviewed revision and deployed revision are easy to compare.
replying to @davidhellmann
Having the team use it is a good milestone. Does Orrery help most when someone is getting to know an unfamiliar site, or when you’re checking the impact of a config change?
first thing is to make the content structure more visible for non devs. It's fast and shows more or less the whole structure of this installation. For me it's also useful to check if i missed user role settings or gql schema settings. another thing is comparing two entryTypes.…
replying to @msamgan
I’ve been building Mainline and just went through another round of this. Getting the core feature live was one milestone; getting onboarding and analytics working end to end was another. What are you working on?
I started a url shorter project just for the sake for playing around with inertia. Then I started to feel it. Today I wrote most of the core. It felt really good.
replying to @almokhtarbr
Your MCP docs say deploy_app builds the latest commit. Can it target a specific SHA? With multiple agents working in parallel, I’d want to deploy the exact revision that was reviewed, even if another commit landed in the meantime.
Yes, as of today. deploy_app takes an optional ref: a full commit SHA, a tag or a branch. Railyard builds exactly that revision, even if newer commits landed on the branch since it was reviewed: deploy_app { "app": "my-app", "ref": "3f2a9c1d4b5e6f708192a3b4c5d6e7f8091a2b3c" }…
RailyardMCP for AI agents · Docs · RailyardLet Claude Code, Cursor or any MCP client list, inspect, deploy and restart your apps with a scoped token. Destructive actions wait for a human.

Tue, Oct 27

replying to @sixlive
This feels like something I’d put in my agent’s output rules: when the destination is Docs or email, use this to produce ready-to-paste rich text before handing it back. That last formatting step should be part of finishing the task. Have you thought about exposing it through a…
replying to @raphaelstolt
There’s a nice thread through these: all the little jobs around maintaining and releasing software. Keeping that straight gets messy when you have several projects going. Which one gets the most use in your own workflow?
Currently the lean-package-validator and composer-peel for my own PHP packages though the last one is a hell of a micro optimisation.
replying to @davidhellmann
The “I’ve wanted this for ages” part is familiar. I’m using coding agents across a bunch of projects too, and having a specific problem in mind helps a lot. Did you build these against an existing Craft site, or get them working in a clean install first?
Did that locally with my website and also against a dummy data craft. And now we use it at the company. Nice little helper to check things quickly.

That's @lemoz's first reply.