Our Research and Software
At Making Made Easy LLC, we focus on one thing: proving that the hardest problems in AI and robotics can be solved with creative engineering.
We don’t rely on supercomputers or million-dollar labs. We show what’s possible with small machines, smart code, and obsession.
Major Achievements
Overload Project
We developed the first public method to run the full, unquantized Mistral-7B language model on as little as 2 GB of RAM. No quantization. No GPU. Just a breakthrough in execution that streams model weights layer by layer, keeping memory use tiny while delivering full-scale LLM performance. Overload proves that cutting-edge AI doesn’t need expensive hardware — just smart design.
ChatGPT/AI-Powered Robot
We built a real, working robot that doesn’t just follow scripts — it sees, hears, and talks via large language models. Our open-source ChatGPT/AI Robot combines sensors, a camera, and voice interaction with AI decision-making. The result: a robot that can hold natural conversations and respond intelligently to the world around it. What was once science fiction is now a buildable project anyone can replicate.
Function Former
We created a self-healing autocoder for Python. Starting from a plain-language goal from the user, Function Former interviews you for specifics based on the goal you give it, writes a structured outline, reviews/revises the outline, then generates the program section-by-section, reviewing/revising as it creates. Once the full script is assembled, it runs a pre-run QA sweep (imports, globals, required functions, and a valid main() guard), executes the program, and—if anything breaks—reads the traceback and patches only the smallest target (a single function, imports/globals, or main) before re-running. It works with either ChatGPT or offline local models (currently: Mistral 7b-Instruct v0.3, Qwen2.5 14b-Instruct, and Qwen2.5 32b-Instruct), all inside an easy-to-understand GUI.
Other Software From Making Made Easy
Tiered-DCA Crypto Trading Bot — An autonomous Robinhood Crypto trader that ladder-buys into dips with configurable tiers (e.g., −2%/−4%/−8%/−16%/−32%), enforces a per-asset max allocation, updates cost basis after every fill, and auto-takes profit once a position is ≥5% green.
Why It Matters
Most people think advanced AI and robotics require huge budgets, powerful servers, or entire research labs. Our work proves otherwise. We’ve shown that a Raspberry Pi, a budget laptop, or a tiny microcontroller can achieve what many consider impossible.
This matters because it democratizes AI and robotics. By publishing everything open-source, we give individuals, small teams, and businesses access to breakthroughs that would normally be out of reach. We’re not just tinkering — we’re breaking barriers and sharing the results.
Making Made Easy is about turning the impossible into something practical, buildable, and open. We push boundaries, release our work, and invite others to take it further. What others call impossible, we call a challenge.