MelTuc Weekly — Edition #2 | April 6–9, 2026
This was the week the factory stopped being theoretical. In four days, I went from zero deployed apps to a portfolio of five, shipped a blog review, and started tracking 169 GitHub repos — all built and orchestrated through the autonomous agent pipeline I've been quietly assembling. The highlight? Watching apps I didn't manually code go live at real URLs and actually work.
Category: Reference
Every factory needs a blueprint. Example Blueprint is the design system reference and component showcase that underpins everything else I'm building. It's the source of truth for UI patterns, typography, spacing, and component behavior across the entire MelTuc portfolio. If you've ever tried to maintain visual consistency across multiple apps without one of these, you know exactly why I built this first. Think of it as the internal style guide that the agents reference when spinning up new apps — it keeps things from going off the rails aesthetically.
Category: Marketing
The mothership is live. MelTuc.com is the main marketing site for the company — what I do, how the factory works, and where the blog lives. It's the public face of everything. Getting this out the door in week one felt important: I wanted a real home for the work before I started talking about it publicly. The blog already has its first post up, which you'll see below.
Category: Marketplace
Okay, this one is a little personal. Pokemon GO Deal Hunter is an automated eBay deal hunter built specifically for Pokemon GO assets — accounts, items, bundles, whatever surfaces. The agent monitors listings, applies deal logic, and surfaces the best finds. I play the game, I've watched the secondary market for it, and I knew there was a real use case here. It's also a good stress test for the marketplace-monitoring pattern I want to use in other contexts. If it works for Pokemon GO, it works for a lot of things.
Category: Intelligence
The agentic AI space moves fast — uncomfortably fast if you're trying to keep up manually. Agentic Intelligence Feed is a multi-source content aggregator and intelligence layer built specifically for this space. It pulls from multiple sources, surfaces what's signal versus noise, and gives me (and anyone else using it) a cleaner view of what's actually happening in agentic AI. I built this partly for my own sanity and partly because I think there's a real gap between "AI news" and "what practitioners actually need to know." This is my attempt to close that gap.
Category: Developer Tools
GitHub Trends Tracker does exactly what it sounds like — it watches GitHub for trending repos, tracks velocity over time, and mirrors selected repos to GitLab. The velocity tracking is the part I care about most: it's not just about what's trending today, it's about which repos are accelerating and which ones are fading. This week it's already tracking 169 repos, which fed directly into the blog review below. The GitLab mirroring is a bonus feature I added for redundancy and offline access to repos I care about.
repo: ray-project/ray · Python · ⭐ 42,029 · Trend: Rising
Ray is genuinely impressive — a mature distributed compute framework that can scale Python and ML workloads from a laptop to a full cluster. But with 3,500+ open issues and a surface area that keeps expanding, I wanted to write something more honest than the usual hype. The full review breaks down exactly which parts of Ray are worth your attention and which parts to approach with caution — worth a read before you commit to it in production.
📦 New apps deployed 5
✍️ Blog reviews published 1
📝 Reviews in draft 0
🔭 New repos tracked 169
🗂️ Total portfolio apps 5
📊 Total repos tracked 169
💡 BRI ideas generated 0
I want to be honest about what this week actually felt like: chaotic and exciting in equal measure. Shipping five apps in four days sounds clean in a newsletter, but in practice it meant a lot of debugging agent outputs, a lot of "why is this URL not resolving," and at least one moment where I questioned whether the whole factory idea was going to hold together under real conditions. It did — but barely, and with friction I'm already working to reduce.
The thing I keep coming back to is that the factory model only works if I stay disciplined about what I'm building and why. The Pokemon GO Deal Hunter is a personal project that doubles as a pattern test. The Agentic Intelligence Feed is infrastructure for my own research. The GitHub Trends Tracker feeds the blog. These aren't random apps — they're supposed to compound on each other. Whether that flywheel actually spins the way I'm imagining is something I'll know more about in a few weeks.
Next week I want to get the BRI (Build-Review-Iterate) pipeline generating ideas consistently — you'll notice that counter is sitting at zero, which means the ideation layer isn't fully wired up yet. That's the priority. If the factory is going to scale, it needs to be generating its own backlog, not waiting on me to come up with the next thing to build. More on that soon.
MelTuc Weekly is published most weeks by a solo developer who builds production web apps using an autonomous AI agent factory. If someone forwarded this to you, you can find the archive at meltuc.com.