nelson

On-chain Researcher | Web3 Writer | I pray that you eventually win 🥇

[nelson @nelsnworie]

I’ve been thinking about what happens after someone finishes a task on @axisrobotics.

Collecting a trajectory is only the first step.

That run still has to get processed, checked, and prepared for training. Then the model gets tested on the same kind of task.

When it fails, that failure becomes the next brief. You don’t just collect more of the same. You collect around the part that broke: the slip, the bad angle, the object that didn’t sit how the model expected.

Collect → train → test → find the failure → collect better data.

Axis isn’t only trying to grow a dataset. It’s trying to run a loop where each round of data can make the next round more useful.

That’s also why more contributors matter. More people means more tasks, more rooms, more edge cases. Variety is what gives the model more physical situations to learn from.

This is how I understand what Axis is building: not just a place to record robot data, but a system that can keep feeding better data into the next training round.

Start contributing here: https://t.co/2Gba1iXQJI

@axisrobotics $AXIS

[nelson @nelsnworie]

@Web3Enchantress @fartkmmonsolana Reminder set

[nelson @nelsnworie]

@Web3Enchantress Happy Wednesday

[Yeminiz 💙🫵 @yeminiz]

good morning friends

@NeoSoulAI x @ActionModelAI campaign ends in 16 hours

remember there’s a $5,000 prize pool to be shared

simply complete the NeoSoul tasks through the Action Model Browser Extension.

how to start:
● open ActionModel at using the link below 👇
● navigate to ActionFi
● choose https://t.co/mHxUyAoYwo on the ActionFi Projects list
● create your first AI agent on NeoSoul

easy to navigate

[nelson @nelsnworie]

@yeminiz @NeoSoulAI @ActionModelAI Appreciate the reminder

[nelson @nelsnworie]

@yeminiz @Bitbase_Ex iphone 18 is out yet?

[Yeminiz 💙🫵 @yeminiz]

another day to wake up to a new iphone

good know that someone is waking up to a new iPhone every morning for the next two weeks.

@Bitbase_Ex pulls a random trader off the leaderboard at 4am UTC daily. not the top ranked one. Random. so being active is the whole requirement.

$300K USDT pool running underneath it.
ends September 18.

might as well be you.

get in now: https://t.co/aLueagn7Ro

[nelson @nelsnworie]

@yeminiz @axisrobotics I use mouse in combination with screen touch. Practice more will master

[nelson @nelsnworie]

I’ve been thinking about what happens after someone finishes a task on @axisrobotics.

Collecting a trajectory is only the first step.

That run still has to get processed, checked, and prepared for training. Then the model gets tested on the same kind of task.

When it fails, that failure becomes the next brief. You don’t just collect more of the same. You collect around the part that broke: the slip, the bad angle, the object that didn’t sit how the model expected.

Collect → train → test → find the failure → collect better data.

Axis isn’t only trying to grow a dataset. It’s trying to run a loop where each round of data can make the next round more useful.

That’s also why more contributors matter. More people means more tasks, more rooms, more edge cases. Variety is what gives the model more physical situations to learn from.

This is how I understand what Axis is building: not just a place to record robot data, but a system that can keep feeding better data into the next training round.

Start contributing here: https://t.co/2Gba1iXQJI

@axisrobotics $AXIS

[nelson @nelsnworie]

@Maria_rare0101 @axisrobotics Very impressive

[nelson @nelsnworie]

@dee_e6 @axisrobotics Did you use my ref link?

[nelson @nelsnworie]

@Web3stunner_ @axisrobotics Very real advantage

[nelson @nelsnworie]

@NKLinhzk @axisrobotics Honestly

[nelson @nelsnworie]

@Jife790 @axisrobotics Agree

[nelson @nelsnworie]

@Akanimo_dx @axisrobotics Correct

[nelson @nelsnworie]

@Fayyonchain @axisrobotics Yes exactly

[nelson @nelsnworie]

@Chocofweb3 @axisrobotics Yes this is actually impressive

[nelson @nelsnworie]

@peanutt @axisrobotics You can participate in the training, they may be reward for participation

[nelson @nelsnworie]

@TeaonaX Gm Tea

[nelson @nelsnworie]

I’ve been thinking about what happens after someone finishes a task on @axisrobotics.

Collecting a trajectory is only the first step.

That run still has to get processed, checked, and prepared for training. Then the model gets tested on the same kind of task.

When it fails, that failure becomes the next brief. You don’t just collect more of the same. You collect around the part that broke: the slip, the bad angle, the object that didn’t sit how the model expected.

Collect → train → test → find the failure → collect better data.

Axis isn’t only trying to grow a dataset. It’s trying to run a loop where each round of data can make the next round more useful.

That’s also why more contributors matter. More people means more tasks, more rooms, more edge cases. Variety is what gives the model more physical situations to learn from.

This is how I understand what Axis is building: not just a place to record robot data, but a system that can keep feeding better data into the next training round.

Start contributing here: https://t.co/2Gba1iXQJI

@axisrobotics $AXIS

[nelson @nelsnworie]

@Xerxes_sama Hope to get better daily

[Xerxes ⧉ @Xerxes_sama]

take out time to work on yourself today.

read a book.

hit the gym.

go for a walk or drive.

set some goals.

be fkn delusional.

gm and stay super productive.