🧠🪰 Could a real fly brain control a drone?

Abbos Aliboev — Thu, 17 Sep 2026 20:06:00 +0000

A few weeks ago, I came across something interesting: Google and HHMI Janelia had mapped the entire brain of a fruit fly - around 166,700 neurons and 125 million connections.

As someone who works with computer vision, I wondered:

Could we use a real brain map to control something?

This isn’t really my field, but I found the idea interesting and decided to try it.

I built a virtual fly-like drone that sees through its camera and moves based on signals from the fly brain. I also created a live dashboard to visualize the brain, camera, and drone movements in real time.

Then I turned it into a small game: Can the fly find the apple? 🍎

The result: yes, it found it.

However, the system still needs help from FlyVis, a pre-trained brain model, plus my own processing and control logic. So the connectome alone is not enough yet.

Why is this interesting?
A fly’s brain uses very little energy, yet it can fly, avoid obstacles, and find food. Understanding how its brain works could help us build smaller, cheaper, and more energy-efficient robots and drones in the future.

I’m continuing the experiment and sharing what works, what doesn’t, and what I learn along the way.

More coming soon. 🚀

abbosaliboev / fly-brain-drone

A real 166,700-neuron fruit-fly connectome piloting a virtual drone to find food. Work in progress, honestly documented.

🧠🪰 Fly Brain Drone

We connected a real fruit-fly brain map to a virtual drone and asked it to find an apple.

166,700 real neurons. 125 million real synapses. One (very) hungry drone.

A real neuroscience dataset. A real physics engine. A real, honestly-reported mess of a research project — in public, as it happens.

An actual screenshot of the live dashboard — not a mockup. Left: the arena from outside. Center: what the drone itself sees. Top right: 47,261 real MaleCNS neuron positions with 27,523 visual-type neurons glowing by live activity. Bottom right: the hexagonal compound-eye input and motor/telemetry readout.

What is this?

In 2026, HHMI Janelia, Cambridge, and Google Research published the first complete connectome of an adult fly's central nervous system — every one of 166,700 neurons and ~125,000,000 synapses in a male Drosophila brain and nerve cord, mapped from real electron-microscope images.

We asked a simple…

View on GitHub

I spent 200 hours building the ultimate free AI resource list so you don't have to (2026)

Abbos Aliboev — Fri, 17 Apr 2026 04:26:25 +0000

I'm going to save you hundreds of hours and potentially thousands of dollars.

Over the past few months, I've been building something quietly: a verified, weekly-updated directory of every legitimately free AI tool, API, and service available to developers in 2026.

Not "free trial." Not "free with a credit card." Actually free.

Here's what I found.

The Dirty Secret About AI Costs

Big Tech's marketing works. Most developers I talk to assume they need to pay $20-$200/month for AI capabilities in their apps. They reach for OpenAI's API, spin up AWS infrastructure, and subscribe to Midjourney.

They don't have to.

What's Actually Free in 2026

Free LLM APIs (No Credit Card)

Provider	Model	Free Limit
Groq	Llama 3.3 70B	14,400 req/day
Google AI Studio	Gemini 2.0 Flash	1,500 req/day + 1M context
Mistral	Mistral 7B	500M tokens/month
Cohere	Command R+	1,000 req/month
Together AI	50+ models	$5 free credit
Cerebras	Llama 3.1 70B	30 req/min (ultra-fast)

Groq alone gives you 14,400 free requests per day. That's a full production app for most indie projects — completely free.

Free Image Generation

Leonardo.ai — 150 images/day, photorealistic quality, zero credit card
Stable Diffusion (local) — unlimited, runs on your laptop
Ideogram — 10/day, best for text-in-image
Kling AI / Hailuo AI — video generation, free daily credits

Free AI Hosting

HuggingFace Spaces — host any Gradio/Streamlit app for free, permanently
Modal — $30/month free GPU compute (serverless)
Cloudflare Workers AI — 100K inference requests/day at the edge
Google Colab — free T4 GPU for notebooks

Local Tools That Beat Paid Alternatives

Ollama — run Llama, Mistral, Phi locally with one command
Open WebUI — self-hosted ChatGPT UI, connects to any LLM
Whisper (local) — transcription that beats paid APIs at $0/month
ComfyUI — image generation workflow tool, unlimited

Get Running in 60 Seconds

Free LLM (cloud):

pip install groq
# Get free key at console.groq.com — no CC
python -c "
from groq import Groq
client = Groq(api_key='YOUR_KEY')
r = client.chat.completions.create(
    model='llama-3.3-70b-versatile',
    messages=[{'role':'user','content':'Hello!'}]
)
print(r.choices[0].message.content)
"

Free LLM (local, fully offline):

curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3.2
ollama run llama3.2

How Much Can You Save?

Use Case	Paid Stack	Free Stack	Annual Savings
LLM API	$150/mo (GPT-4o)	$0 (Groq)	$1,800
Image gen	$50/mo (Midjourney)	$0 (Leonardo)	$600
Hosting	$200/mo (AWS)	$0 (HuggingFace)	$2,400
Transcription	$40/mo	$0 (local Whisper)	$480
Total	$440/mo	$0	$5,280/yr

The Full List

I've compiled 700+ of these into an open-source repo that I update every week:

👉 github.com/abbosaliboev/free-ai-bible

It includes:

Complete tables for every category above
One-click setup scripts (run one command, everything configured)
Copy-paste app templates: chatbot, RAG pipeline, image generation API, voice assistant
Comparison charts: which LLM to use, local vs cloud, image gen picker
Translations in 6 languages (Chinese, Korean, Spanish, Russian, Japanese, Uzbek)

Zero affiliate links. Zero sponsored entries. If it's there, I tested it.

If this saves you money or time, a ⭐ star on the repo is all I ask. It takes one second and helps other developers find it.

What free AI tool has surprised you the most? Drop it in the comments — the best ones will make it into the list.

DEV Community: Abbos Aliboev