Duende: Web UX for guiding Gemini as it improves your source code
I wrote a simple web UX in Python/JavaScript that spawns a conversation with Google Gemini with MCP commands that let it work on a specific coding task that you specify: http://github.com/alefore/duende The UX lets you observe the conversation and provide guidance (e.g. "don't implement Foo through Bar, that's suboptimal; instead …"). It supports a `--review` mode, where once the main conversation says "I'm done with the task", various "evaluation" conversations are spawned, each focusing on reviewing the change from a very specific angle (e.g. "does it introduce…
What it does
In the maker’s words, at launch
I wrote a simple web UX in Python/JavaScript that spawns a conversation with Google Gemini with MCP commands that let it work on a specific coding task that you specify: http://github.com/alefore/duende The UX lets you observe the conversation and provide guidance (e.g. "don't implement Foo through Bar, that's suboptimal; instead …"). It supports a `--review` mode, where once the main conversation says "I'm done with the task", various "evaluation" conversations are spawned, each focusing on reviewing the change from a very specific angle (e.g. "does it introduce useless comments?"). In the future, I'm considering adding other workflows. I've used it mostly to develop itself (I started with a super rudimentary manual implementation and then mostly used Duende to extend its own implementation) as well as (with moderate success) to add a few new features to [my C++ text editor](http://github.com/alefore/edge). It's been a lot of fun. I'm still developing my intuitions for what works and what doesn't, but I've already had plenty of experiences where I've been WOW'ed by what LLMs can already accomplish (as well as, to be honest, plenty of very disappointing cases where it struggles significantly on tasks that I expected would have been trivial). It's a learning experience, seeing things like how to avoid hallucinations; developing an intuition for how much to break down large tasks into smaller ones; or knowing when to abort a conversation and restart it with an improved prompt (vs continue to steer it in the right direction). I think investing up-front in setting up a good context (e.g., good validation logic, useful constant contexts, a good set of "review evaluators" that can push-back against sloppy code) can go a long way to increase the odds of success. It has changed my perception on the applicability of AI for developing software. While I've reviewed all the outputs (and often rewrite parts of them manually), I'm already incorporating it into (some parts) of my development life-cycle. I have many ideas for improvements, but figured I'd share this early and ask for feedback. Hopefully others find it interesting; would love to hear your thoughts.
Does the same job
all alternatives →
Open Source AI Chrome ExtensionSep 2025 · ▲87Chrome extension allow the user to communicate with the AI
More dev tools this month
the category →
Open-source GTM skills for technical founders
Dev tools · 29d ago · gtmcofounder.com
OpenTrailPaper is open-source bike computer firmware for the LilyGO T5S3 4.7" E-Paper PRO. It supports offline maps, GPX routes, FIT recording and Bluetooth sensors.
Dev tools · 2d ago · opentrailpaper.com
Launched alongside, July 2025
the whole month →
Trickle - Magic Canvas ▲1,064The 1st Agentic Canvas for building apps visually with AI
AI · 2025 · trickle.so
- TB
Life & fun · 2025 · trytender.app
- TY
Today marks ten years, 3653 consecutive days, of running at least one mile every day under the USRSA rules [1]. To celebrate, I built an interactive dashboard that turns a decade of GPX files into charts you can explore. Running has truly changed my life: I've made lifelong friends, explored beautiful places, and more importantly invested into my own health and fitness, which I'm starting to see the positive benefits as I get older. The stack is pretty simple: a NextJS app, with a Postgres database to keep all my running data, and all the stats are pre-computed and cached in Redis, so I…
Dev tools · 2025 · nodaysoff.run
- DA
Made this website as an exercise in vibe-coding and GCP. It was posted about a few times around the internet, on sites like Morning Brew (https://www.morningbrew.com/issues/business-buzzkill), MetaFilter (https://www.metafilter.com/209703/Draw-A-Fish), boingboing.net, etc. I think it's cute! I built a basic CNN trained against penises and swastikas, and then anything that doesn't hit the 63% confidence score gets sent to a mod queue, a vibe-coded fish-tinder…
Dev tools · 2025 · drawafish.com
