Paul Colagrande

92 posts

Paul Colagrande banner
Paul Colagrande

Paul Colagrande

@ColagrandePaul

Kaspa Ambassador Dev student at Epitech Paris Passionate about blockchain & DAG technology

Paris, France Katılım Aralık 2023
66 Takip Edilen414 Takipçiler
Sabitlenmiş Tweet
Paul Colagrande
Paul Colagrande@ColagrandePaul·
I’m thrilled to officially join the #Kaskad team after six inspiring weeks of collaboration! Thank you to everyone for the warm welcome and for the opportunity to develop the Oracle. Excited to keep pushing the boundaries together. 🚀
Kaskad@AppKaskad

After 6 weeks challenging @ColagrandePaul on the modelling of the algorithm of our Oracle with @eliottmea, we are excited to share that he has exceeded expectations. Today we are officially welcoming him to the #Kaskad team, where he can continue putting his talents to work. Welcome to the crew, Paul!

English
4
10
122
4.2K
Paul Colagrande retweetledi
eliott
eliott@eliottmea·
Have spent two weeks annoyed at an attack vector on my HFT oracle, nicely pointed out by someone in the #kaspa discord. Specifically on illiquid exchanges. will detail this later, but essentially there should be a way to make it such that people keeping their #kas (or later other assets) on CEXs, have a win-win situation. we distribute software (essentially automating arb bots for people on a specific CEX), so that anyone can 1. profit if someone tries to manipulate the oracle (one specific attack in mind, ie a huge arbitrage). 2. help secure the oracle through filling this arbitrage. this probably won't be a constant profit because we'll frontrun you on small arbs, but if someone tries to attack the oracle, the whole community becomes rich. Y'all in ?
English
12
18
114
4.9K
Chewie
Chewie@Chewie_xo·
The @kaspaunchained community goes hard! Great to see so many people attend the 4th anniversary celebration with people flying in from all around the world.
Chewie tweet media
English
10
8
78
1.3K
Paul Colagrande
Paul Colagrande@ColagrandePaul·
@KaspaLens Also, does your backend implement any kind of oracle or pricing mechanism? And for the orderbook view, is it sourced live from exchanges or reconstructed on your side? Thanks in advance for the clarification.
English
1
0
3
75
Paul Colagrande
Paul Colagrande@ColagrandePaul·
@KaspaLens Great work, the tool looks impressive! Quick question regarding the price displayed on kaspa-lens.com: what’s the source of your KAS price feed? Is it aggregated across multiple exchanges, or taken from a single market?
English
2
0
3
111
KasLens
KasLens@KaspaLens·
This detailed view into the order books of exchanges for $KAS is sometimes pretty fascinating. We’re one step closer to releasing this tool. We’re currently tackling the last performance issues, showing a highly detailed, clustered view requires a lot of resources, and we’re implementing additional technical solutions to make the experience as smooth as possible. Keep in mind, we’re talking about potentially millions of elements being drawn in the browser. We’ve already switched to WebGL, but that isn’t always enough. Stay tuned, kaspa-lens.com will bring this future to the $KAS community for free. 💚
KasLens tweet media
English
7
29
129
4.1K
Paul Colagrande retweetledi
Kaskad
Kaskad@AppKaskad·
Unveiling our Fundraising Campaign through the $KSKD token A lot has been brewing behind the scenes, and it is time we start opening the doors. 📣Behold! The official Kaskad Campaign starts today: a full month of learning, discovery and community participation. It will be the first time we share the full picture of the protocol, including the role of the $KSKD token in the platform, and how the community can take part in the project’s next chapter. At #Kaskad, we do not take education lightly. In fact ,we believe that educating our participants at this stage is an absolute need. Here is what to expect by clicking on the link below: • Building the best lending platform on Kaspa - our approach and long-term vision • Understanding the Kaskad tokenomic.- interactive tools and guided explanations • Community Fundraising - a chance to take part in the early growth phase • Deep dive into Kaskad - interviews, events, technical sessions and Q&A • ...and a few more surprises! Kaskad is built by Kaspa lovers, for Kaspa lovers, which is why our fundraising campaign will be open to communities and DAO VCs only. Ready? Let the campaign begin! Rendez-vous here: kaskad.app/fundraising Invitation code: KASKADR1
English
10
23
106
9.9K
Paul Colagrande retweetledi
3SEARCH Capital
3SEARCH Capital@3searchcapital·
We’re thrilled to announce our strategic partnership with @AppKaskad - the first compliant lending protocol on #Kaspa 🔒 Their expertise in oracles, verified price feeds, and institutional-grade infra brings DeFi one step closer to Kaspa’s future.
3SEARCH Capital tweet media
English
3
16
54
2.1K
Paul Colagrande retweetledi
eliott
eliott@eliottmea·
Coming to @AppKaskad (and more generally to #kas) tldr; oracle inspire from the NA (no arbitrage) from math fin theory (which obv doesn't hold). Idea : if there were a single LOB across the market, oracles wouldn't make any sense. We need them only because of fragmented liquidity. Thus at a fixed time t, we aggregate snapshots from all CEX LOBs (assume for now kas only trades on CEXs). We use websockets such that, when an LOB updates (an order hits/cancels in the book), we receive the info sub 50ms or faster. Given there is natural arbitrage (no perfect market), the oracle simulates arbitrage in order to understand what the fair value is : the value above (resp. under) which there is net excess supply (resp. demand). Maybe our engine will try to fill the arb itself, didn't think about that yet. Ok we got fair value, is our oracle robust ? No. Although illegal, nothing prevents a CEX from manually overriding their API and displaying #KAS/USDT=0.000001, as well as setting a huge sell wall at 0.000002 with volume larger than all buy orders across venues. We let the reader understand how this would successfully manipulate the oracle. Two outlier detection methods that only work well when used together. The first, using survival analysis in order to detect if an arbitrage is lasting longer than usual (conditioned on market conditions such as volatility). Eg. usually an arb of 2% between MEXC and kraken is filled within 1s, weirdly it's taking 5s and the arb is still here : why is nobody taking the free money ? Hence we observe something we'll call epsilon-events (V_t being the volume of arb at time t): Problem w. the above, the same attack mentioned still works, because there is a huge arb (by a CEX creating fake sell wall), but the oracle will necessarily take a positive amount of time to flag the outlier. In that window, it will account for it and get wrecked. Second detection : Wasserstein-1 distance between exchanges. An LOB histogram can be seen as a pile of sand, and the distance from one lob to the other is the amount of energy (or trades) needed to move one pile of sand to the next. If for eg., in 1 update, kraken reports a change that we deem too large (has to be VERY carefully defined), we flag it as an outlier. The oracle should then attempt to profit from that HUGE arb opportunity, and if it doesn't manage to within eg. 50ms, successful flag. If it does manage, we profit huge money, we're/I'm rich and unflag. Now the only thing that needs to be really studied, apart from what parameters we use (will need @remao155 ML skills for this), as well as what can happen under collusion. Although imo collusion won't change much. challenge the idea and find some flaws cc. @maxibitcat, @emdin, @xximpod, @quex_tech, @madenis, @JC_G33K etc. @Kaspa_KEF will send report soon, on this as well as my work with yonni (both related anyways), @ColagrandePaul there is much work ahead, pls implement all the above ;)
eliott tweet media
English
12
23
95
6.2K
Paul Colagrande
Paul Colagrande@ColagrandePaul·
Congrats @remao155 ! Thrilled to see you joining the Kaskad team, your background and research focus sound like a perfect fit. Excited to build together!
Kaskad@AppKaskad

We are glad to welcome Rémi Khani @remao155 as an Applied Researcher at Kaskad! He will be focusing on the tokenomics game theory behind Kaskad’s protocol. Rémi has recently graduated with a Master’s degree in Applied Mathematics and Machine Learning from ETH Zurich and holds a Bachelor’s degree in Pure Mathematics from EPFL Lausanne. Currently in the process of writing a research paper, he is conducting a proof of concept for the protocol’s tokenomics under @BagayokoJack supervision, exploring the mathematical foundations of its key features. Building on Kaskad's tokenomic model, Rémi is analyzing optimal incentive structures to promote sustainable TVL growth, long-term participation, and responsible leverage, all aimed at maintaining the platform’s stability and efficiency. Together with @eliottmea , our Lead Oracle Architect, he will also be responsible for developing the Research and Development Department at Kaskad. Rémi’s previous work experience in machine learning ranges from medical imaging for stroke detection to early decision making in time series environment using reinforcement learning models. Strongly passionate about DeFi and Kaspa, he is keen on finding a way to contribute to the Kaspa ecosystem by joining our team. Please give that man a follow and welcome to #Kaskad, Rémi! 🚀 ‌

English
0
0
7
486
KRCBot
KRCBot@KrcBot·
Yes, kafka and elastic are fine. DruidDB is also good for your existing stack. alt test. 1M messages per second publishing = Rabbit Stream + NATS Jetstream Columnar stores and TSDB data - at 1Billion+ records with sub 1 second queries - QuestDB is worth a test to compare/contrast.
English
1
0
2
46
KRCBot
KRCBot@KrcBot·
Oracles drive pricing within a tight variance, enabling rebates and AMM trades. When a major exchange with deep liquidity (e.g., Binance) sees a sharp move, others (e.g., KuCoin) adjust rapidly to avoid arbitrage losses via API trades. Exchanges prioritize protecting their USDT reserves over your token, favoring large buy-ins and trapped orders over sell-offs to prevent liquidity drains.
English
1
5
19
1.7K
Paul Colagrande
Paul Colagrande@ColagrandePaul·
@KrcBot @eliottmea We process all data in real time, then store both the order books and computed results at T+1 after processing. Currently evaluating Kafka and Elastic for scalable storage and analytics.
English
1
0
1
44
KRCBot
KRCBot@KrcBot·
@ColagrandePaul @eliottmea Yes, this would be the way. For the websockets are you doing 1-1 or are you pushing to a pub/sub system? For the storage, to do historical lookbacks you're using something faster than timescale/influx hopefully (like questdb).
English
1
0
0
45
Paul Colagrande
Paul Colagrande@ColagrandePaul·
@eliottmea @KrcBot Thanks for the CCXT link, great resource. We’re using Rust with WebSockets instead of REST to stream L2 order book data live, testing aggregation every 50 ms, and detect subtle market changes with lower latency. Still, CCXT is handy for backtesting and managing API calls.
English
1
0
3
54
Paul Colagrande
Paul Colagrande@ColagrandePaul·
@TheLoadedLounge Heard about your project and got inspired. I built a quick front-end demo. Would you be open to discuss this further in DMs so I can hear more about your ideas for the app?
Paul Colagrande tweet media
English
1
0
3
65
Kaskad
Kaskad@AppKaskad·
Say hello to @ColagrandePaul, @eliottmea's cousin, who just joined #Kaskad as our new junior developer intern. Paul is a 3rd-year CS student at Epitech (FR) with a strong interest in blockchain and decentralized systems. Active in the community as a Kaspa ambassador for the past 9 months, Paul has shown real commitment to contributing and learning. For example, it is following his demand the Rust learning channel was created in the discord. During his part-time internship (Thu/Fri), Paul will focus on building his skills, benefiting from mentorship, and contributing to the growth of the Kaspa ecosystem, with a particular interest in DeFi lending as a cornerstone for future development. We are very pleased to welcome him on board!
Kaskad tweet media
English
11
11
114
19K