Proof of concept: real-time fire detection from an autonomous drone
Live camera → fire detection → annotated evidence → agent decision loop → movement command
@manakoai detects the fire
@chutes_ai decides the next move
@SwarmSubnet turns it into a flying agent
We're not ready for what happens when subnet intelligence enters the physical world
Code tomorrow.
Open-source. Again.
Swarm has locked $2M, representing ~1M ALPHA, following the conviction update
Over 50% of the staked supply will be locked
That includes:
• ~$1M from us, locked perpetually (~90% of our total holdings)
• ~$630k already locked by @const_reborn (100%)
• ~$330k to be locked by @Ever_BlueGreen (100%)
We’re all-in, and our conviction has never been higher.
See you at PoT.
We locked our ALPHA with @SwarmSubnet today, now the 3rd largest wallet after the Swarm team and @const_reborn.
Proud to show support not only with our initial investment but also with full conviction lock.
The team behind Swarm matters most to us. We look forward to our physical lab in the Andorran mountains.
#Bittensor
Swarm has locked $2M, representing ~1M ALPHA, following the conviction update
Over 50% of the staked supply will be locked
That includes:
• ~$1M from us, locked perpetually (~90% of our total holdings)
• ~$630k already locked by @const_reborn (100%)
• ~$330k to be locked by @Ever_BlueGreen (100%)
We’re all-in, and our conviction has never been higher.
See you at PoT.
@sama We need robots we can trust.
Permissionless open ownership.
You offer a walled garden.
A closed door. Zero trust.
“Open” robotics will be built and distributively owned on #Bittensor
Starting with @SwarmSubnet
Sim-to-real is the holy grail of robotics.
And one of the first places it breaks is sensors.
In @SwarmSubnet, miners learn in simulation using depth information.
But not every real drone has a depth sensor...
So we’re experimenting with deep-learning-based depth estimation:
📷 real drone camera frames
→ 🧠 estimated depth maps
→ 🧪 compared against simulation depth
The goal is simple:
make simulation closer to what a real drone can actually perceive.
That’s what you’re seeing here.
Step by step, closing the sim-to-real gap.
Thanks for your continued support.
Swarm will also be locking a large portion of the owner alpha this Monday.
Details are still being finalized, and we’ll share more as soon as they’re confirmed.
Everblue & Green welcomes Bittensor’s mainnet upgrade to spec version 411.
With conviction locks now delivering immediate power and healthier decaying mechanics, we will be locking 165k in alpha with @SwarmSubnet.
We believe in this subnet’s value and the real-world problems it solves, but far more, we believe in the founding team behind it.
Good teams build real solutions that last.
It's Friday, perfect time for a miners perf update.
Ever wondered if the models our miners distil are actually good enough to use in the real world?
Well, @swarmsubnet just took our 'detect fire' model and dropped it inside a drone. And it works beautifully.
Detect fire is one of the skills live on the network right now, and the miners are still pushing it higher.
Watch below.
The full perf update is in the next tweet 👇
Everblue & Green welcomes Bittensor’s mainnet upgrade to spec version 411.
With conviction locks now delivering immediate power and healthier decaying mechanics, we will be locking 165k in alpha with @SwarmSubnet.
We believe in this subnet’s value and the real-world problems it solves, but far more, we believe in the founding team behind it.
Good teams build real solutions that last.
The repo includes:
- live Tello camera streaming
- Score fire model integration
- annotated frame + video recording
- Chutes command-selection loop
- dry-run mode before flight
Built as a small, inspectable PoC.
Link in the next tweet.
bittensor:native
Proof of concept: real-time fire detection from an autonomous drone
Live camera → fire detection → annotated evidence → agent decision loop → movement command
@manakoai detects the fire
@chutes_ai decides the next move
@SwarmSubnet turns it into a flying agent
We're not ready for what happens when subnet intelligence enters the physical world
Code tomorrow.
Open-source. Again.
Proof of concept: real-time fire detection from an autonomous drone
Live camera → fire detection → annotated evidence → agent decision loop → movement command
@manakoai detects the fire
@chutes_ai decides the next move
@SwarmSubnet turns it into a flying agent
We're not ready for what happens when subnet intelligence enters the physical world
Code tomorrow.
Open-source. Again.
DJI, we need to talk.
It’s not you, it’s us.
We need to become expert pilots before we can develop the autopilot.
We might buy a few drones to start automating some things.
SWARM is going all-in
Our mission has always been to unlock real-world autonomy for drones. Now we are accelerating.
We are moving beyond one challenge at a time — autopilot, search and rescue, navigation — and expanding toward challenge families: a broader framework to train autonomous flying agents across many missions, environments, and real-world constraints.
We already rented our physical lab and are actively building it out. This is where we will close the sim-to-real gap.
We will also announce a new CTO: a robotics expert with deep experience in autonomous systems. She will lead the technical side, from drone development to real-world testing.
For miners: starting with the next challenge, we will reduce burn to 80%, effectively making emissions 4x more impactful than today.
Huge thanks to EBG for supporting this decision. No more little-by-little.
We are building the lab.
We are bringing in robotics leadership.
We are expanding the benchmark.
We are going after general autonomous flying agents.
The next phase of SWARM starts now.
Sneak peak of the office attached BTW
DJI, we need to talk.
It’s not you, it’s us.
We need to become expert pilots before we can develop the autopilot.
We might buy a few drones to start automating some things.
SWARM is going all-in
Our mission has always been to unlock real-world autonomy for drones. Now we are accelerating.
We are moving beyond one challenge at a time — autopilot, search and rescue, navigation — and expanding toward challenge families: a broader framework to train autonomous flying agents across many missions, environments, and real-world constraints.
We already rented our physical lab and are actively building it out. This is where we will close the sim-to-real gap.
We will also announce a new CTO: a robotics expert with deep experience in autonomous systems. She will lead the technical side, from drone development to real-world testing.
For miners: starting with the next challenge, we will reduce burn to 80%, effectively making emissions 4x more impactful than today.
Huge thanks to EBG for supporting this decision. No more little-by-little.
We are building the lab.
We are bringing in robotics leadership.
We are expanding the benchmark.
We are going after general autonomous flying agents.
The next phase of SWARM starts now.
Sneak peak of the office attached BTW