Let people know that you can't rot in vain, you can become the sun and shine.✨💙
Joined April 2023
- Tweets14.2K
- Following2.5K
- Followers1.7K
- Likes40.2K
Pinned Tweet
Gm all 💙
Small daily actions, big compounding. Say GM, read the latest 2Factor Brief, and watch your Marks grow with the Buffalo multiplier. The early grind is where Season 1 gets won. 🦬
points.2factor.finance/r/2es…
@2FactorFinance
🤝 Paid partnership · 🤖 Made with AI
Getting my @miuadotfun Genesis Pass for Early Access
Bonding curve NFT marketplace on Robinhood Chain & Arc.
Register here: miua.fun
Tomorrow end! teapaw.org/miuadotfun/status/2119…
🤝 Paid partnership
mahb00b.base.eth retweeted
@primus_labs 태스크 영상 가이드예요 쉽고 수수료도 적어요
1 링크로 들어가 확장 프로그램이 없으면 설치 페이지가 열려요
2 저는 예시로 github를 눌렀고 나머지도 비슷해요
3 지갑 트랜잭션을 승인해요
4 github 페이지에서 확장 프로그램이 자동 진행해 포인트를 받아요
s.kaito.ai/t8XwBgd
🤝 Paid partnership
mahb00b.base.eth retweeted
GN Friends 😍
Before going to bed take a look at this
Me in Prisma #17 😍
The IROS demo did not run on the arm that collected its training data
That detail is easy to skip
@PrismaXai fine-tuned π0.5 on YAM pick-and-place, about 668 episodes, roughly 10 hours
Then they ran the checkpoint on a different YAM rig, in a hall nobody had trained in
Same model family
Different hardware
Different room
People walking behind it
🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻
That is a harder test than a clean replay on the original table
A policy can look fine when the cameras, the joint offsets, and the lighting match the training setup
Move the arm, change the background, shift the object by a few centimeters, and the success rate tells you what the dataset actually generalized
This is why the last posts kept circling the same filter
✅A completed task is not the dataset
✅A scored episode is not the dataset
✅Even a clean fine-tune is not the dataset, if it only works on the rig that produced it
🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻
The useful question is whether the behavior survives a different arm and a room it has never seen
That is what they took to the floor
That is also what they are taking to CoRL in Austin, November 9 to 12
@MaxC16134 @vivianrobotics @AmberShiroo @tamaraa901 @Mim5100
GN Family
Me in Prisma #16 😍
A physical arm cannot run a thousand tests before lunch
That is why this repo exists
@PrismaXai open-sourced yamulator, a path-traced simulator for the YAM arm
They built it to run thousands of π0.5 rollouts a day without resetting a real scene
That is the practical version of the Agility point from the last post
You do not want every bad idea to touch hardware Hardware breaks
Hardware takes weeks
A sim lets you throw the bad checkpoints away before they ever see a gripper
They used it for the fine-tune behind the IROS demo
About 10 hours of YAM pick-and-place, 668 episodes, then large-scale evaluation in the sim instead of on the table
The data blend stays private
The training settings and the sim do not
If you have a YAM, you can run the same kind of ablation on your own rig
The next public demo is CoRL in Austin, November 9 to 12
Different room, same question
Does the policy still hold when nobody trained in that hall
--------------------------------------------------------
For more info check these links below:
Discord:
🔗discord.gg/ngG7qPNaA
The website:
🔗 prismax.ai
@MaxC16134
@vivianrobotics
@AmberShiroo
mahb00b.base.eth retweeted
Justice is supposed to protect people, not become another burden they have to fight to survive.
But look at these stories.
» A case that takes 75 years.
» A man forced into bankruptcy over a debt he never owed.
» Another man spending nearly 30 years on death row for crimes he didn't commit.
And the painful part?
These aren't just failures of the system. They are people's lives being destroyed while the system takes its time.
This is why I believe justice needs an upgrade.
» Justice shouldn't depend on how much money you have to hire a lawyer, how long you can afford to fight, or whether powerful institutions are willing to admit they were wrong.
The problem isn't that justice is expensive. It's that the cost of getting justice can sometimes become greater than the injustice itself.
» We need systems that are faster, more transparent, and accessible to ordinary people. Systems where evidence matters more than authority, where decisions can be challenged, and where accountability isn't optional.
Technology alone won't solve everything, but it can help us rethink how justice works.
Because justice that arrives decades too late, or only for those who can afford it, isn't really serving the people it's meant to protect.
Justice shouldn't be a privilege. It should be something everyone can access.
#JusticeNeedsAnUpgrade
@GenLayer
Slow. Expensive. Wrong.
It's been this way for centuries. It doesn't have to be.
#JusticeNeedsAnUpgrade
mahb00b.base.eth retweeted
Gfungo 💛
Privacy in crypto is usually treated as one feature. What caught my attention about @FungoLabs is that it shows up at three different stages of the same journey, and each stage normally leaks something someone else can exploit.
1. At mint: the surprise comes back
On a transparent chain, traits and rarity are readable almost instantly. Rarity tools rank every mint in seconds, snipers move first, and the "not knowing" part of minting disappears.
With FHE from @zama, each Genesis NFT starts with encrypted traits. The art can be revealed when the holder chooses, while the Essence (class, stats, element) stays encrypted on chain. Nobody can pre read rarity, because there is nothing to read.
2. At auction: bids that can't be watched
2,500 of the 3,232 Genesis NFTs go through an encrypted sealed-bid auction. Because bids stay hidden, there's nothing to monitor and outbid at the last second, and every winner pays the same clearing price. That shifts the game from reacting to others toward deciding what the NFT is worth to you.
3. At claim: rules written in code
Per the project, the $FUNGO claim window is 7 days, the deadline is hardcoded in the contract, and unclaimed tokens are burned instead of piling up for later decisions. There's also an option to claim as confidential cFUNGO, so the amount doesn't have to be visible to everyone.
Why this combination is interesting
Each stage has the same weakness on a normal chain: information that's public too early gets used against the participant. Encrypting it, and putting the rules in code, moves the advantage away from whoever reads the data fastest.
What I'm still watching
→ How FHE scales with larger media and heavier computation
→ The cost and speed of encrypted operations
→ Whether the launchpad and games get built on top, since that's where hidden traits become real utility
A privacy feature only matters when people use it for something they couldn't do before. Sealed bids and hidden trait gameplay might be exactly that.
If you could keep one thing private in an NFT, what would it be: the art, the stats, or the price you paid?
fungolabs.org/launchpad/c/cf
🤝 Paid partnership · 🤖 Made with AI
💛 ZK can prove something about hidden data. FHE goes a step further: a smart contract can actually compute on data while it stays encrypted.
That's what @FungoLabs builds on, using FHE from @zama.
Each Genesis NFT has two encrypted layers:
→ The Look: the artwork. The holder sees it first and decides whether to make it public.
→ The Essence: class, stats and element. These stay encrypted on-chain, but contracts can still use them.
Why it matters:
→ Traits are encrypted at mint, so rarity can't be pre-read or sniped
→ 2,500 of the 3,232 Genesis NFTs go through an encrypted
sealed-bid auction, so nobody can watch competing bids and all winners pay the same clearing price
→ Hidden traits can power games without exposing a player's stats to opponents
Still worth watching: how FHE scales with large media, the cost of encrypted operations, and whether the launchpad and games actually get built.
Which matters more in an NFT: showing the art, or protecting what's behind it?
Launchpad: fungolabs.org/launchpad/c/cf
Gfungo 💛
Privacy in crypto is usually treated as one feature. What caught my attention about @FungoLabs is that it shows up at three different stages of the same journey, and each stage normally leaks something someone else can exploit.
1. At mint: the surprise comes back
On a transparent chain, traits and rarity are readable almost instantly. Rarity tools rank every mint in seconds, snipers move first, and the "not knowing" part of minting disappears.
With FHE from @zama, each Genesis NFT starts with encrypted traits. The art can be revealed when the holder chooses, while the Essence (class, stats, element) stays encrypted on chain. Nobody can pre read rarity, because there is nothing to read.
2. At auction: bids that can't be watched
2,500 of the 3,232 Genesis NFTs go through an encrypted sealed-bid auction. Because bids stay hidden, there's nothing to monitor and outbid at the last second, and every winner pays the same clearing price. That shifts the game from reacting to others toward deciding what the NFT is worth to you.
3. At claim: rules written in code
Per the project, the $FUNGO claim window is 7 days, the deadline is hardcoded in the contract, and unclaimed tokens are burned instead of piling up for later decisions. There's also an option to claim as confidential cFUNGO, so the amount doesn't have to be visible to everyone.
Why this combination is interesting
Each stage has the same weakness on a normal chain: information that's public too early gets used against the participant. Encrypting it, and putting the rules in code, moves the advantage away from whoever reads the data fastest.
What I'm still watching
→ How FHE scales with larger media and heavier computation
→ The cost and speed of encrypted operations
→ Whether the launchpad and games get built on top, since that's where hidden traits become real utility
A privacy feature only matters when people use it for something they couldn't do before. Sealed bids and hidden trait gameplay might be exactly that.
If you could keep one thing private in an NFT, what would it be: the art, the stats, or the price you paid?
fungolabs.org/launchpad/c/cf
🤝 Paid partnership · 🤖 Made with AI
💛 ZK can prove something about hidden data. FHE goes a step further: a smart contract can actually compute on data while it stays encrypted.
That's what @FungoLabs builds on, using FHE from @zama.
Each Genesis NFT has two encrypted layers:
→ The Look: the artwork. The holder sees it first and decides whether to make it public.
→ The Essence: class, stats and element. These stay encrypted on-chain, but contracts can still use them.
Why it matters:
→ Traits are encrypted at mint, so rarity can't be pre-read or sniped
→ 2,500 of the 3,232 Genesis NFTs go through an encrypted
sealed-bid auction, so nobody can watch competing bids and all winners pay the same clearing price
→ Hidden traits can power games without exposing a player's stats to opponents
Still worth watching: how FHE scales with large media, the cost of encrypted operations, and whether the launchpad and games actually get built.
Which matters more in an NFT: showing the art, or protecting what's behind it?
Launchpad: fungolabs.org/launchpad/c/cf
💛 ZK can prove something about hidden data. FHE goes a step further: a smart contract can actually compute on data while it stays encrypted.
That's what @FungoLabs builds on, using FHE from @zama.
Each Genesis NFT has two encrypted layers:
→ The Look: the artwork. The holder sees it first and decides whether to make it public.
→ The Essence: class, stats and element. These stay encrypted on-chain, but contracts can still use them.
Why it matters:
→ Traits are encrypted at mint, so rarity can't be pre-read or sniped
→ 2,500 of the 3,232 Genesis NFTs go through an encrypted
sealed-bid auction, so nobody can watch competing bids and all winners pay the same clearing price
→ Hidden traits can power games without exposing a player's stats to opponents
Still worth watching: how FHE scales with large media, the cost of encrypted operations, and whether the launchpad and games actually get built.
Which matters more in an NFT: showing the art, or protecting what's behind it?
Launchpad: fungolabs.org/launchpad/c/cf
🤝 Paid partnership · 🤖 Made with AI
This quoted post is unavailable.
Good morning fam💙
✍️ Dostoevsky writes:
The greatest punishment for a person is to be imprisoned in his past. Every time you say (if I hadn't done that...) you are blackmailing the past. The truth is that the past is a creditor who never settles accounts; the more you think about it, the more it will swallow up your share today
The only way to settle with the past is to declare bankruptcy:
Accept that the copy of you made a mistake, pay the price and close the case. 👍🏻
Have a nice day.🤗
mahb00b.base.eth retweeted
Building the future of private DeFi. 🛡️
Rift DEX Beta is officially live on Seismic Testnet! Experience confidential swaps & concentrated liquidity today.
🔗 Link in bio to test the dApp & faucet.
Drop your feedback below! 👇
Gmic lovely fam 🤎🩷🩵
Have a great day
Tnx for good ideas bro That brings the community closer together and creates a better sense of participation. @Ravondir🌷
In the end, I hope that the project will be good for all of us.
@SeismicSys
@xealistt
@NoxxW3
🤝 Paid partnership · 🤖 Made with AI
Gmic Family🫶
I made a Seismic card featuring Rocky welcoming members of our community at the airport!
Want your own card? Send me a DM on Discord (Ravondirx) or Twitter with your PFP, username, and Magnitude.
Let’s help Rocky welcome every one of us. That way, every Seismic community member can have their own little story in this collection.
@SeismicSys
@NoxxW3 @xealistt
Try to be one of the early here, my friend.😊👍🏻
We’ve raised $500,000 in an Echo round, backed by 68+ angel operators & Web3 creators.
We’re building ONLORE to connect emerging projects with the creators who help them grow.
Gaming at the core. Open to what’s next.
Write the lore. Earn the access.
onlore.io
Maybe it's time for the Treasure Hunters. @onlorehq ✌️
Every world starts with a story.
ONLORE Season 1 is live with a $100,000 + 10% $LOR prize pool (planned).
Discover new worlds.
Write the lore.
Earn Season Shards.
Climb the leaderboard.
Your Shards decide your share.
The first 300 to verify become Founding Lorekeepers ✦
mahb00b.base.eth retweeted
Me in Prisma #15 😍
SunDay TacoDay 🌮 let's start today with @PrismaXai
A university lab can have a paper idea and still have no robot data
That is the gap this update actually targets
@PrismaXai opened a University Research Program
Selected labs get free access to the Robotics Data bank
If the task is not already in the catalog, they collect it to the lab’s spec
✓ No equity
✓ No exclusivity
✓ No claim on the model the lab trains
What they ask for is a citation
Train on it, publish on it, release the checkpoint, even share the episodes as a benchmark
☆ Just cite the source
That is a different door from the commercial plans
🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻
Discover and Researcher are paid access to the existing bank
Enterprise is custom collection for teams with a budget
This one is for labs that have a thesis and not a data team
The hard part was never “can we collect a run”
It was collecting runs worth training on, then getting them in front of people who will actually train
A university lab usually cannot staff a teleop fleet
It can still publish on clean data if someone else collected and scored it
For more info you can check the link below:
🔗prismax.ai/university-resear…
-------------------------------------------------------
Also check these links below:
Discord:
🔗 discord.com/invite/prismaxai
The website:
🔗 prismax.ai/
@MaxC16134 @vivianrobotics
@AmberShiroo @Mim5100
Me in Prisma #14 😍
One line from the latest RoboCon panel stuck more than the event recap
Vivek from Agility put it simply:
If you are testing a policy on a hundred-thousand-dollar humanoid and it breaks, that is it
The team cannot evaluate the next policy until the robot is repaired
That can take months 😳
So the first place to test is simulation
Not because sim replaces the real robot
He was clear that the sim-to-real gap is real
❌If you only evaluate in sim and deploy zero-shot, it fails
What sim does is narrow down which policies are worth the risk of running on hardware
That is the same logic as the data posts, just one layer earlier
🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻
✅A noisy teleop run should not become training data
✅A weak policy should not become a hardware test
Both waste the expensive part:
Real contact with the world
@castorhat said @PrismaXai tests the same way
• Failures in sim are trivial to recover from
• Failures on hardware cost weeks
Sim also works as a regression suite
Every past failure becomes a test case
You do not want to run 10,000 real-world rollouts for every release
------------------------------------------------------
For more info Check these links below:
Discord:
🔗discord.gg/ngG7qPNaA
The website:
🔗 prismax.ai
@MaxC16134 @vivianrobotics @AmberShiroo
Most leverage is built for people who leave.
Perps liquidate. Leveraged ETFs decay.
@2FactorFinance is built for people who hold: Kelly-optimized ~1.33x BTC exposure, no liquidations, live in beta on Base.
Leverage for holders, not traders.points.2factor.finance/r/0eh…
🤝 Paid partnership
جی پریسمااااااااااااا😻
🏆 کوییز توپ طلای پرشین
⚽
آمادهاید اطلاعات فوتبالیتون رو به چالش بکشید؟
قرارِ کنار هم لحظات فوتبالی پرهیجانی رو رقم
بزنیم! 🔥
🎙 با میزبانی هاستهای برنامه:
👤 @AvinEsmaelpour
👤 @Mim5100
فوتبالیها کجان؟ پرچم کی
بالاست؟ 🚩👀
توی این رقابت داغ منتظرتون
هستیم!🩵🩷
🗓 زمان: دوشنبه
⏰ ساعت: 5 عصر
📍 @IR مکان
@PrismaXai
🤝 Paid partnership · 🤖 Made with AI
🏆 کوییز توپ طلای پرشین
⚽
آمادهاید اطلاعات فوتبالیتون رو به چالش بکشید؟
قرارِ کنار هم لحظات فوتبالی پرهیجانی رو رقم
بزنیم! 🔥
🎙 با میزبانی هاستهای برنامه:
👤 @AvinEsmaelpour
👤 @am32016
فوتبالیها کجان؟ پرچم کی
بالاست؟ 🚩👀
توی این رقابت داغ منتظرتون
هستیم!🩵🩷
🗓 زمان: دوشنبه
⏰ ساعت: 5 عصر
📍 @IR مکان
There's a new version of this post
mahb00b.base.eth retweeted
Genm
Season 2 of @GenLayer has officially started, and we're all at the starting line.🏁
I know a lot of us are still waiting to hear about the TGE. Mochi seems to have a sealed envelope in hand, but isn't giving anything away just yet.
But while we're waiting for that envelope to open, we shouldn't let the track sit empty. There's still plenty to explore, learn, and do in Season 2.
So why not get moving now? That way, when the news finally drops, we'll already be a few steps ahead.
🤝 Paid partnership
mahb00b.base.eth retweeted
کریپتو از صفر
ـــــــــــــــــــــــــــ
اگه تازه وارد دنیای کریپتو شدی، این سری برای توئه، مفاهیم ساده و قدم به قدم
#کریپتوـازـصفر
🤖 Made with AI
mahb00b.base.eth retweeted
Me in Prisma #13 😍
The useful @PrismaXai update this week was not another event post
They published how the π0.5 demo from IROS was actually built
One detail fits the last posts better than the booth photo did
> A Python pipe buffer was holding camera frames until 64 KB piled up
> That added hundreds of milliseconds of lag
> After they fixed it and cut inference from 230 ms to 87 ms, the arm started moving more like their operators
That is the same loop, just from the model side
🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻🔻
✅If the operator is slow or jerky, the dataset inherits it
✅If the rollout itself is laggy, the policy starts looking laggy too, even when the data was clean.
The other notes are worth reading if you care about the stack:
● Under hard lighting, training only the action expert lost about half the success. The full fine-tune, which also trains the VLM, did not
● Two memory fixes dropped peak VRAM from about 22 GB to 11.2 GB, so it runs on 16 GB cards
● The sim and rollout repos are open: yamulator and prism-pi
The demo itself ran on a different YAM rig than the one its training data came from
Site-unseen was not only the room
It was the hardware too
Next public stop is CoRL in Austin, November 9 to 12, booth 1100
Stay Tuned with PrismaX 😍
-------------------------------------------------
If you want to know more about Prisma
Check these links below:
Discord:
🔗 discord.gg/ngG7qPNaA
The website:
🔗 prismax.ai
And if you want not to miss anything save this: 🔗linktr.ee/PrismaX_AI
@MaxC16134 @vivianrobotics @AmberShiroo
@Mim5100
Me in Prisma #12 😍
The IROS recap from @PrismaXai landed on Oct 6th and one detail was more interesting than the booth photo
Their demo was a π0.5 checkpoint, zero shot, on two YAM arms
The hall was new
People were walking behind the robot all day
Nobody had trained in that room before the doors opened
🔽🔽🔽🔽🔽🔽🔽🔽🔽
That is the part that connects to the last posts
A scored episode can look clean
A live room does not care
Light changes, someone passes behind the arm, the object is not where the dataset expected
Site unseen is a harsher test than "task completed"
The other note from the floor fits the same problem
A lot of hand makers were not only selling hands
They were selling capture gloves and exoskeletons next to them
A lab can buy a few hands
It still cannot collect enough demonstrations on its own
The hand is the easy sale
The data is the hard part
Next stop they mentioned is CoRL in Austin, November 9 to 12, with a new demo
----------------------------------------------------
If you want to know more about this event Dont miss
@PrismaXai latest posts 😍
Follow this #IROS2026
Also check these links below:
Discord:
🔗 discord.gg/ngG7qPNaA
The website:
🔗 prismax.ai
And if you want not to miss anything save this: 🔗linktr.ee/PrismaX_AI
@jonstephens85 @ieeeiros
@MaxC16134 @vivianrobotics @AmberShiroo
