"I remember reading this exact bug 18 months ago in a @code4rena contest. I cannot, for the life of me find it again."
Auditors don't have a search problem. they have a recall problem.
Scar fixes that.
R@10=0.901 over 232k smart contracts at 114ms.
live demo at https://t.co/Kfr142IDAi.
🚨 OFFICIAL: Lionel Messi, TOP SCORER EVER in World Cup history joint with Miro Klose. 🎞️⭐️
16 goals at the World Cup. 💥🇦🇷
In one night, on 6th World Cup debut game at 38, Messi surpasses Mbappé (14), Müller (14) and Ronaldo Nazario (15).
LI.FI Intents is live, yes finally!
This is a huge huge moment for LI.FI
First things first, big congrats to @jimchang@stabiliseret and the Intents & Solver squad for shipping LI.FIIntents
architecturally, it's hands down the best solution in the market - open, modular, and customizable in every way, built alongside the great teams that contributed to the Open Intents Framework
every part of the system is well thought out and built with the different stakeholders in mind
over the last few weeks I've personally talked to all the solver teams that are live filling order flow on LI.FI Intents, and all of them were excited to see how the system plays out - with different elements in place like the reputational system that makes competition fairer between solvers without compromising on reliability or execution quality
beyond excited to see how the system performs at scale, with billions in order flow moving through
Secondly, LI.FI Intents is an integral part of LI.FI's product suite as we push to get more institutions, neobanks and fintech companies to build products onchain
the three use cases we're focused on come from customer demand we've heard across current and future clients
Stablecoin payments
Access to real-world assets
Access to compliant liquidity
All three are critical for putting crypto into the hands of millions of users, without the UX feeling like 'crypto' at all - they don't need to know
the primitives our industry has built can stand on their own merits - technologically superior to anything the world of finance has seen, and more capital efficient than anything traditional rails offer today
We don't even need to go into the benefits of blockchain rails like transparency and everything being public - those are the icing on the cake
What we've built is better money technology
And with LI.FI Intents, we now have a product our team can push to embed crypto into every financial product in the world
Big things ahead, excited to be locked in and building with this team.
PS: the Intents launch will forever be remembered internally for this epic shoot we all got to work on together in Crete!
Instead of celebrating this I'm resolving issues of mates on-boarding to web3.
Everything is broken all talks about being *Decentralized* but your entry starts with a centralized exchange giving away your credentials.
Imagine getting your first usdt without giving credentials or getting scammed.
Building the first end-to-end smart contract audit stack where every layer is grounded in retrieval at @ScarRetreival
The keystone is SCAR — a sparse code retriever I built that hits R@10 = 0.901 retrieving from a 232k-contract corpus in 114ms.
BM25 collapses to 0.308 at that scale.
What that unlocks across the stack:
Layer 1 (next) — agent harness calls into the tool layer
Layer 2 (built) — SCAR + reranker + context-aware MCP, exposed to any agent (Claude/GPT/Qwen) over MCP
Layer 3 (training data built, RL env next) — GRPO with verifiable rewards: the retriever grades the model Layer 4 (next) — Qwen3.6-27B, SFT → GRPO against SCAR
The novel bit: SAE-LoRA — low-rank adaptation of Sparse Autoencoder encoder weights. 4.6M params (0.3% of the backbone) turns reconstruction features into retrieval-discriminative ones. 37.6× over the frozen SAE baseline.
Weights, training pairs, eval set all open-source.
Building the first end-to-end smart contract audit stack where every layer is grounded in retrieval at @ScarRetreival
The keystone is SCAR — a sparse code retriever I built that hits R@10 = 0.901 retrieving from a 232k-contract corpus in 114ms.
BM25 collapses to 0.308 at that scale.
What that unlocks across the stack:
Layer 1 (next) — agent harness calls into the tool layer
Layer 2 (built) — SCAR + reranker + context-aware MCP, exposed to any agent (Claude/GPT/Qwen) over MCP
Layer 3 (training data built, RL env next) — GRPO with verifiable rewards: the retriever grades the model Layer 4 (next) — Qwen3.6-27B, SFT → GRPO against SCAR
The novel bit: SAE-LoRA — low-rank adaptation of Sparse Autoencoder encoder weights. 4.6M params (0.3% of the backbone) turns reconstruction features into retrieval-discriminative ones. 37.6× over the frozen SAE baseline.
Weights, training pairs, eval set all open-source.
"I remember reading this exact bug 18 months ago in a @code4rena contest. I cannot, for the life of me find it again."
Auditors don't have a search problem. they have a recall problem.
Scar fixes that.
R@10=0.901 over 232k smart contracts at 114ms.
live demo at https://t.co/Kfr142IDAi.
Instead of celebrating this I'm resolving issues of mates on-boarding to web3.
Everything is broken all talks about being *Decentralized* but your entry starts with a centralized exchange giving away your credentials.
Imagine getting your first usdt without giving credentials or getting scammed.
UX is still so much broken in web3 or I'm not updated with current meta?
Ex-Classmates got excited that I'm launching my research into a product and they wanna donate for the cause but cex->p2p->wallet->verify->donate.
Onchain ramping is so much broken with absurd quotes.
Never got the opportunity to work with C4.
I mailed them for using their data for @ScarRetreival but never got the revert back.
And I don't think thats a case now left, bummer.
After careful consideration, we’ve made the decision to wind down @code4rena. This community has meant a great deal to everyone who has been part of building it, and sharing this news is not easy.
Your Jumper Pass unlocks up to 65% travel discounts and cashbacks.
You also earn extra XP for booking stays on Jumper Travel.
Book and earn XP: https://t.co/Nod6XnAuSO
Smart contract security gets stronger when past findings are easy to reuse.
@ScarRetreival by @LifeHodlr is an open-source AI retrieval system that helps auditors and AI tools quickly surface relevant vulnerability precedents from 232,000 real Ethereum contracts.
Learn more: https://t.co/yADATrvBzf
Support: https://t.co/DvPNJUG1dW
Won $400 at the @lifiprotocol DeFi Mullet Hackathon — developer track. Adding $100 of my own.
The full $500 → @Giveth Ethereum Security QF round.
https://t.co/l5hJbpMM1F
$10 to each of 50 projects securing this ecosystem.
Thanks @PhilippZentner@brucexu_eth@kenny_io@arjunnchand
Won $400 at the @lifiprotocol DeFi Mullet Hackathon — developer track. Adding $100 of my own.
The full $500 → @Giveth Ethereum Security QF round.
https://t.co/l5hJbpMM1F
$10 to each of 50 projects securing this ecosystem.
Thanks @PhilippZentner@brucexu_eth@kenny_io@arjunnchand
For full transparency — I'm also a project in this round.
@ScarRetreival — sparse retriever for smart contract audits. R@10 = 0.901 on 232K Ethereum contracts. Apache 2.0.
The 50 donations stand alone. This is about the ecosystem 🛡️
https://t.co/FM4XDuPaZz
Technical report:
https://t.co/wstw4FsZk0
"I remember reading this exact bug 18 months ago in a @code4rena contest. I cannot, for the life of me find it again."
Auditors don't have a search problem. they have a recall problem.
Scar fixes that.
R@10=0.901 over 232k smart contracts at 114ms.
live demo at https://t.co/Kfr142IDAi.
Less “someone should fund open-source security”
More “I just did” 🫡
QF math: 100 people donating $10 each beats one whale dropping $500.
500$ just got 12,920$ matching to 50 projects!
$1M+ pool. 134 projects. ~60 hours left.
If you’ve been waiting on this round… it’s now.
https://t.co/l5hJbpMM1F
This is important.
When you donate, do it to multiple projects.
That makes capital allocation more effective and that makes your donation more IMPACTFUL, because it has a unique fingerprint, therefore it gives more signal, therefore it counts more.