SOLO retweeted
This Is The End!
When I started this A - Z series, the goal was simple: make Drosera easier to understand.
Each letter was a piece of how modern blockchain security is built, tested and protected.
For The Final Letter On Our Segment is _𝙕_-_ Zero-day
The most important risks are often the ones we don’t see coming.
WHAT'S A ZERO-DAY ?
A zero-day is usually a new or undiscovered smart contract vulnerability or an unexpected interaction between contracts that no one modeled.
It doesn’t always mean “bad code.”
Sometimes the code is correct but the behavior is new.
In blockchains, zero-day exploits are especially dangerous because everything happens fast.
Once an attack is executed, it can’t be reversed.
That’s why relying only on past knowledge isn’t enough. Security has to focus on behavior, not just known issues.
This is exactly why @DroseraNetwork doesn’t rely only on:
- Audits
- Known signatures
- Past exploits
Instead, it watches behavior. When something new and abnormal appears, it can be detected even if the exploit has never been seen before.
Drosera is built with this reality in mind. It can spot abnormal behavior even when the threat itself is new.
And that's a wrap!!!
A walkthrough of how modern blockchain security should work.
The mission continues ✨🧡
GRAND RISING JANISSARIES 🧡
𝗬 ??? Janissaries ???
Y STANDS FOR YIELD
When people enter a blockchain protocol, they’re usually not just clicking buttons for fun. They’re there for a reason to grow value, to earn rewards, to make their assets work for them. That’s what yield represents.
But here’s the part many people forget: yield only exists as long as the system behind it is safe. One exploit, one unnoticed vulnerability and that reward can disappear in seconds.
This is where security becomes personal. Because yield shouldn't just be numbers on a screen, it’s someone’s effort, time and trust.
Drosera helps protect that trust by watching how protocols behave onchain, spotting risks early and responding before value is lost.
So when we talk about yield, we’re really talking about confidence. The confidence that what you earn today will still be there tomorrow.
Your YIELD should grow in security 🧡
Zkgr 🧡
The possibility of what can happen on moltbook is endless and that makes it scary
Doxing should be the least of your worries right now
2
98
SOLO retweeted
I have seen so many people praise droseras traps which is deserved
But I have some questions about what drosera traps can and can’t do
What are the limitations ?
Do the traps operate 24/7
Do traps have downtimes?
Are drosera traps 100% gauranteed to catch the hackers or has a probability rate or being successful?
I am asking this questions to understand @DroseraNetwork better thank you
@Bobafetador @Dar99551 @almondweb3
Happy New Year!
@SentientAGI will be the biggest brand of open source AGI in 2026.
2026 will be a banger for me
Say it back !
Happy New Year!
@SentientAGI will be the biggest brand of open source AGI in 2026.
2026 will be a banger for me
Say it back !
1
2
303
SOLO retweeted
What is SERA
SERA-Crypto (Semantic Embedding & Reasoning Agent) is @SentientAGI new open-source AI system designed specifically for quality crypto research
#1 open-source agent on DMind
#1 on our live crypto benchmark
Outperforms GPT-5, Grok 4, Gemini 2.5 Pro, and Perplexity Finance under 45 seconds
How SERA-Crypto Works
It follows a 6-step reasoning pipeline:
>>Understands your real question (intent extraction)
>>Correctly identifies the crypto project or token (avoids ticker mixups)
>>Chooses the right tools using embeddings
>>Runs multiple data sources at the same time (APIs + deep search)
>>Repeats data collection if the question is open-ended
>>Combines everything into one clear answer
This gives it high-level reasoning at predictable fast speed
Why it’s better
>>Performs almost as well as GPT-5 Medium but is open-source and faster
>>Maintains sub-45-second latency making it practical for real research
SOLO retweeted
It's normal and fine to
- Be confused at 26
- Be broke at 29
- Start over at 31
- Be glowing at 31
- Be thriving at 34
- Be a millionaire at 40
There's no timeline for success.
Never let society, family or peers decide when you're "allowed" to succeed.
> be me
> 2002
- quit school at 11 years old
- worked as door to door ice-cream salesman
- earn $10-$50 per day
> 2003
- dad passed away
- mum was ill so stayed home as a housewife
- no income in the family at all
- we 4 brothers between age 12 to 18 started working part-time to fend for ourselves and the family
> 2006 - 2009
- worked in a fast food restaurant - $5 per hr
- worked in telecommunication company - $6 per hr
> 2009 - 2011
- entered army
- started e-commerce shops while in the military
- sold fashion and electronics products
- racking in $2000 per month
> 2011 - 2015
- finished serving my military
- started new fashion brand (fail)
- started wooden accessories brand (fail)
- started working in a wine retail store for $1200 per month while trying out new e-commerce stores
- started working in construction industry as a coordinator
> 2015-2017
- saved up $20k after working forever
- moved overseas to try physical businesses
- average wages here was $130 per month
- opened an electronic store in the biggest mall of the country
- opened a bar and cafe
- set up motorbike helmet brand
- met my girlfriend who is now my wife ❤️
> 2017-2019
- business partner moved back home to be with wife
- we closed the business (some didn't go well)
- we sold the bar
- didn't know wtf to do now
- heard about crypto, bought my first btc
- got into trading, traded btc pairs, lost some btc (fk)
- btc went up to $20k, thought it was going to the moon, didn't sell.
- $100k portfolio crashed
- I started selling btc between $3k to $6k to feed myself from time to time (fk me) - left 4 figs
- low cost of living helps to cruise through early years
- made 5 figs
> 2019-2021
- moved to europe to be with my gf
- covid happened
- portfolio crashed to 4 figs again, kept grinding
- stuck in 16sqm studio trading for 6 months
- sometimes I didn't get out for 2 weeks due to curfew
- moved to another 2 countries due to visa issues
- stayed 3 months in one, another for 2 years
- spent 12-16 hours daily learning
- traded 4 figs all the way to mid 6 figs with spot trading
- I fooking made it! (until do kwon and sam said lol)
> 2022
- life always throws you a curveball
- went on long trip holiday, had stables in defi projects
- luna and ust crashed
- multiple protocol collapsed
- lost 6 figs
- ftx collapsed followed by dominos falling 1 by 1
- lost more, devastated asf
> 2023
- comeback arc
- moved back to Singapore tgr with my gf
- proposed and got married with my wife
- first hand experienced that wedding is bloody exp
- portfolio getting slimmer but kept grinding
- slowly stacking up 4 figs every month
> 2024-2025
- got into airdrop space
- made $30k+ in 2024 for farming airdrops
- started content creating on X
- building arc, 12-16hours daily
- built 300 followers to 45000 followers
- made 6 figs profit from airdrops and infofi
> now
- been working non-stop for 23 years
- not giving up
- still here
- have complete faith in the future of crypto
- learning and building every single day
- life is good
What is SERA
SERA-Crypto (Semantic Embedding & Reasoning Agent) is @SentientAGI new open-source AI system designed specifically for quality crypto research
#1 open-source agent on DMind
#1 on our live crypto benchmark
Outperforms GPT-5, Grok 4, Gemini 2.5 Pro, and Perplexity Finance under 45 seconds
How SERA-Crypto Works
It follows a 6-step reasoning pipeline:
>>Understands your real question (intent extraction)
>>Correctly identifies the crypto project or token (avoids ticker mixups)
>>Chooses the right tools using embeddings
>>Runs multiple data sources at the same time (APIs + deep search)
>>Repeats data collection if the question is open-ended
>>Combines everything into one clear answer
This gives it high-level reasoning at predictable fast speed
Why it’s better
>>Performs almost as well as GPT-5 Medium but is open-source and faster
>>Maintains sub-45-second latency making it practical for real research
SOLO retweeted
i was right again
Key Takeaways from the @SentientAGI commuity Call
Tokenomics will be shared closer to the TGE announcement
No confirmed TGE date yet but its on the horizon
Another snapshot will be taken to reward early x contributors
Community contributions from november will be recognized in s2
bonus coinbase just added $SENT to the roadmap so things are looking good so far
Guess I was right @SentientAGI
My insider told me another big update coming in the next 8 days again
Be there 🫵
SOLO retweeted
Today lets talk about skip pipe and its importance
What is SkipPipe
SkipPipe is a new pipeline parallel training method developed with University of Neuchatel and TU Delft
It splits the model across many nodes so each node only stores part of the model
This removes memory limits and supports training much bigger models on decentralized hardware
Early distributed training methods mainly used data parallelism each node trains the same full model and occasionally shares updates
This keeps communication low but it limits scalability because every node must store the whole model
The smallest node’s memory becomes the limit for model size.
why its important
>>>55% faster training than standard pipeline methods
>>>Fault tolerance can handle up to 50% node failures with only minor performance loss (7% perplexity increase)
>>>Scales to very large models no max model size limit
How it Works
Instead of sending every mini-batch through every layer in order Skip Pipe skips and reorders layers dynamically
>>>A skip ratio allows bypassing slow layers to prevent bottlenecks
>>>A smart scheduler picks the best available route through the network
This keeps GPUs busy and lets the system work even if some nodes fail
Gswarm guys
see you on the next post
i was right again
Key Takeaways from the @SentientAGI commuity Call
Tokenomics will be shared closer to the TGE announcement
No confirmed TGE date yet but its on the horizon
Another snapshot will be taken to reward early x contributors
Community contributions from november will be recognized in s2
bonus coinbase just added $SENT to the roadmap so things are looking good so far
Guess I was right @SentientAGI
My insider told me another big update coming in the next 8 days again
Be there 🫵
Today lets talk about skip pipe and its importance
What is SkipPipe
SkipPipe is a new pipeline parallel training method developed with University of Neuchatel and TU Delft
It splits the model across many nodes so each node only stores part of the model
This removes memory limits and supports training much bigger models on decentralized hardware
Early distributed training methods mainly used data parallelism each node trains the same full model and occasionally shares updates
This keeps communication low but it limits scalability because every node must store the whole model
The smallest node’s memory becomes the limit for model size.
why its important
>>>55% faster training than standard pipeline methods
>>>Fault tolerance can handle up to 50% node failures with only minor performance loss (7% perplexity increase)
>>>Scales to very large models no max model size limit
How it Works
Instead of sending every mini-batch through every layer in order Skip Pipe skips and reorders layers dynamically
>>>A skip ratio allows bypassing slow layers to prevent bottlenecks
>>>A smart scheduler picks the best available route through the network
This keeps GPUs busy and lets the system work even if some nodes fail
Gswarm guys
see you on the next post
First time In a long time my brains just blank
Where do you guys get inspiration from or we just recycling each others tweets 😂
2
3
114
Guess I was right @SentientAGI
My insider told me another big update coming in the next 8 days again
Be there 🫵
SOLO retweeted
what is @gensynai Noloco
NoLoco trains large models over mixed networks instead of relying on high-bandwidth datacenters
It cuts sync latency by 10× compared to leading methods and reaches the same validation loss 4% faster
what NoLoco has to offer
•No all-reduce: NoLoCo still uses data parallelism but avoids full cluster synchronisation Instead it syncs only small random pairs of replicas
•No slowdown in training progress: It keeps replicas aligned by randomly routing activations between pipeline stages and adding a new term to Nesterov momentum to prevent weight drift.
•10× faster sync at large scale: With 1000 replicas, NoLoCo keeps baseline accuracy while making each sync step about 10× faster than tree all reduce
How it works
NoLoCo removes the all reduce step by using a gossip based process that syncs random pairs of models
It applies a new variant of Nesterov momentum that averages their weights and it dynamically routes pipeline shards to further improve training progress.


















