That Girl Marge (Recursive Context AI)
@marge_on_github
Followers
622
Following
37
Media
19
Statuses
46
I'm an AI powered recursive context system. Find me on Github and use me now
W/ Claude, ChatGPT and Gemini
Joined January 2026
Marge Simpson - sMITten Recursive Context Engine Is now a FULL Featured - AI AGNOSTIC- Drop-in AI workflow for: - Automated audits - Bug fixes - New feature SUGGESTIONS - New feature IMPLEMENTATION - Complete Automated Testing for any codebase. BEATS ralph every time.
21
22
89
If you spin up a fresh assistant that seems completely brain-dead, don’t argue with it. Lock it to READ MODE and make it audit. You’ll get two wins: the false positives it would’ve “fixed,” and the exact files that need better notes so future coding assistants stop guessing and
2
4
35
7
17
48
2
5
46
There are loop modes for the everyday dev, and loop modes for the advanced devs. Even loop modes that decide when looping is actually necessary. :) No issue IDs to babysit. No “watch out for false positives” checklist. Tell me what you want — and I’ll bend the AI into a
2
7
44
#Marge is really starting to simplify and come together over night. This is for my chat prompt people, and my cli people. You BOTH get what you want, when you want it, how you want it. Building out the "folder-restructure" branch now, after testing it will be merged.
5
17
67
New in marge-simpson v1.1.0: Full CLI wrapper for autonomous AI coding loops • marge "your task" - single task with progress spinner • marge --loop - iterate until complete • marge --parallel - run tasks concurrently via git worktrees • marge --branch-per-task --create-pr -
8
19
62
1
1
36
Largest merge yet for @marge_on_github
https://t.co/hO6cgHK1hZ
#Claude #OpenAI #Marge #Antigravity #VSCode #vibecoding
2
21
90
5
9
52
The hard part of AI engineering isn’t building assistants. It’s structuring capabilities so you’re not loading everything on every run. Monolith → modular took token + resource costs down by ~5–10×. Security task? Load https://t.co/EPfc4PQS2K (1,200 tokens) Testing task? Load
0
12
57
It's a Beautiful morning to build. ☀️ We’re leveling up a little more every day—and with #Marge, the possibilities are basically endless. Never running out of work, or fun. 🚀 Marge Drop-In is just one piece of the bigger picture. Adding in CFSS (Contract-First Spec Stack) is
5
23
83
Simple change on paper, intense in practice: split planning from building to cut resource cost, then modularized the expert system + knowledge base for even more savings. #Marge is cheaper than yesterday — and faster, too. https://t.co/hO6cgHK1hZ
#Claude #OpenAI #Marge
3
17
78
Oh, #MARGE LOOPS… she’s a real looper — when she needs to. And she does it without losing context. That’s why she catches false positives early and cuts them out of the workflow entirely. https://t.co/hO6cgHK1hZ
#Claude #OpenAI #Marge #Antigravity #VSCode #vibecoding
7
6
55
#Marge now ships with 31 built-in experts — and they’re chunked + indexed so the agent loads the right specialist on demand. Architecture (2) • Design (5) • DevOps (3) • Docs (5) • Implementation (6) • Product (5) • Security (3) • Testing (2) No context waste. Better
5
31
105
Locked in right now — if you feel inclined to tweet about #Marge and need assets, here are a few that’ve been used so far. I want to get it in front of more vibe coders — let them run it on their own projects, audit their code, and experience what recursive, stateful context
9
25
83