SKS ✓
@sallubroz · joined 21 Jun 2024
Curious about Web3 Sharing what I discover & think Always learning.
22 369Followers
16 451Following
115 809Posts total
14.3KViews on collected posts
Последние посты
Ищем новые посты…
A little sky, a little freedom,
and a whole lot of peace.
Sometimes, that’s all we need. https://t.co/7oNRMBsC3l
471 views · 57 likes
· 0 reposts · 39 replies
Open on X →
me trusting AI with my insurance claim because it left more receipts than i do 😭
meanwhile @RialoHQ making sure every step has a history
bro even AI gotta show proof of work now lol
605 views · 100 likes
· 1 reposts · 73 replies
Open on X →
Everyday crypto becomes useful when it works for the things you already spend money on.
@joinCero is bringing that idea closer to real life, from groceries and monthly subscriptions to travel and daily expenses.
Instead of keeping digital assets separate from normal spending, h
851 views · 153 likes
· 9 reposts · 129 replies
Open on X →
Good morning
Have a good day https://t.co/Uz1JLQWP6A
1.7K views · 266 likes
· 10 reposts · 251 replies
Open on X →
@sallubroz Sleep tight. Tomorrow’s the next level.
11 views · 2 likes
· 0 reposts · 0 replies
Open on X →
good night everyone 🫶
another day done, another little chapter closed.
rest well and come back stronger tomorrow. https://t.co/4h42AQjEGb
1.7K views · 205 likes
· 8 reposts · 170 replies
Open on X →
@sallubroz better data really does make robots way smarter
25 views · 4 likes
· 0 reposts · 0 replies
Open on X →
10/ That's what makes the contributor side of physical AI exciting for me.
My individual trajectory might look tiny.
Maybe I'm just moving a virtual object from one side of a table to another.
But zoom out.
Thousands of people creating different demonstrations, encountering h
47 views · 7 likes
· 0 reposts · 1 replies
Open on X →
9/ Better training data doesn't magically create a perfect robot.
There are still models, hardware, sensors, testing and many other pieces involved.
But the data shapes what the policy gets the chance to learn.
Give it narrow experience and its understanding can remain narrow.
37 views · 7 likes
· 0 reposts · 0 replies
Open on X →
8/ Think about how much richer this can become at scale.
Instead of learning from one operator repeating one movement, a robot policy could learn from a much wider range of demonstrations, corrections and situations.
The important part isn't collecting data just to make a large
46 views · 9 likes
· 0 reposts · 0 replies
Open on X →
7/ This is where @axisrobotics becomes interesting to me.
Through Axis Hub, contributors can work on simulated robot tasks and create trajectories through their actions.
A trajectory isn't simply:
“The object started here and finished there.”
It captures what happened between
35 views · 10 likes
· 0 reposts · 0 replies
Open on X →
6/ And different people can add different experiences.
Give the same manipulation task to several contributors and they probably won't complete it with exactly the same motion.
Someone takes a shorter path.
Someone approaches from another angle.
Someone makes a mistake and ht
63 views · 9 likes
· 0 reposts · 0 replies
Open on X →
5/ This is where human input becomes powerful.
Humans naturally react to small changes without thinking much about them.
If something slips, we adjust our hand.
If an object isn't where we expected, we change direction.
If the first approach doesn't work, we try another one.
35 views · 8 likes
· 0 reposts · 0 replies
Open on X →
4/ Failure becomes especially interesting here.
Suppose a robot reaches for an object and misses.
At first, that looks like a bad demonstration.
But imagine a human takes control at that exact moment, adjusts the movement and completes the task.
Now there is something https:/
46 views · 9 likes
· 0 reposts · 0 replies
Open on X →
3/ This helped me understand something important.
A huge dataset and a useful dataset aren't automatically the same thing.
Robots need variety in what they experience.
Objects can appear in different positions.
People can complete the same task using different movements. http
59 views · 8 likes
· 0 reposts · 1 replies
Open on X →
2/ Now imagine we want to teach a robot to pick up a bottle.
We give it thousands of demonstrations.
Sounds great.
But there is a problem.
What if almost every demonstration shows the bottle standing in the same place?
The robot may become excellent at that setup without htt
62 views · 8 likes
· 0 reposts · 0 replies
Open on X →
1/ First, what is a robot policy?
I like to think of it as the decision-making system behind the movement.
A robot observes what is around it and has to decide what action should come next.
Where should my arm move?
How should I approach this object?
When should I close the
371 views · 30 likes
· 1 reposts · 4 replies
Open on X →
When we watch a robot smoothly pick something up, open a drawer, or place an object exactly where it belongs, the movement almost looks natural.
But the part we don't see is what interests me more.
Before a robot can make a good decision, it needs useful examples to learn from.
8.2K views · 191 likes
· 20 reposts · 137 replies
Open on X →
На фоне аккаунтов своего размера
20 постов за последние 90 дней рядом с диапазоном 10K–100K подписчиков. доходит до меньшего числа людей, зато вовлекает их куда сильнее.
Медианные просмотры53этот аккаунт924медиана для 10K–100K
Охват, %0.24%этот аккаунт3.62%медиана для 10K–100K
Вовлечённость, %19.24%этот аккаунт1.52%медиана для 10K–100K
| Показатель | Этот аккаунт | Медиана для 10K–100K | Отношение |
|---|---|---|---|
| Медианные просмотры на пост | 53 | 924 | 0.06× |
| Охват (просмотры ÷ подписчики) | 0.24% | 3.62% | 0.07× |
| Вовлечённость | 19.24% | 1.52% | 12.7× |
Другие в этом диапазоне → Сравнить с другим аккаунтом → Как считаются эти ориентиры →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
631 Sep
35
46
37
47
25
1.7K6 Sep
12
7
11
1.7K7 Sep
851
605
471
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
14.29%1 Sep
28.57%
19.57%
18.92%
17.02%
16.00%
22.42%6 Sep
16.67%
28.57%
18.18%
31.75%7 Sep
34.20%
28.76%
20.38%
Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.
What the audience does
Likes55.7%1 087 in total
Reposts2.5%49 in total
Replies41.3%805 in total
Quotes0.3%5 in total
Bookmarks0.2%4 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Похожие аккаунты
something ✓
@something
44.7K followers
· 132 posts
TΞ🅰️〽️UЛłÐ🎱S
@Fer_TeamUnidoS
44.7K followers
· 286.1K posts
Hercules | DeFi ✓
@Hercules_Defi
44.7K followers
· 37.1K posts
Alexsander Silva ✓
@Alexsandercrvg
44.7K followers
· 204.2K posts
Laura Albornoz Pollmann
@lauralbornoz
44.7K followers
· 78.6K posts
Joei ⛓🐶
@joeiwanko
44.7K followers
· 40.6K posts
CIF State
@CIFState
44.7K followers
· 12.2K posts
برادران مارکس بجز مانفرد
@behtarin1760
44.7K followers
· 32.3K posts