Point an AI agent at a 500-file DLT repo all at once, and you get shallow observations: a repeated finding in a pile of false positives.
We run it as a campaign instead. 🧵
A bug class in Uniswap V4 hooks that's easy to miss:
If your hook measures AMM state inside the current tick-spacing range to make an economic decision — say, calculating a dynamic fee — and the swap it governs can traverse adjacent ranges, you have a problem.
The hook's measurement captures depth at one point. The swap executes across many. That gap is exploitable.
If your fee or incentive math reads from a localized snapshot but the trade walks a trajectory, it's not a question of whether it will be exploited, but when.
The measurement is a snapshot. The trade is a trajectory.
For years, I was an open source skeptic. I didn’t think open models would ever close the gap with frontier labs and I was wrong. I’ve moved all my research to open source models. Over the next 3-4 months, I’m putting around $1M into my own mini data centre to train and specialize them. The amount of tokens I have spent on frontier labs have been astronomical for the kind of results I was looking for and it has cost me hundreds of thousands of dollars in short period of time and it’s still restricted in speed and quality which isn’t enough anymore for impactful outcomes. A model doesn't need to be great at everything. A team of specialized models, each trained deeply in its own field and working together, can beat a single generalist. I believe that approach can outperform today's frontier models like Fable and Astra by a wide margin.
With the current advancement in AI I believe the GPU prices will skyrocket even further and soon it’ll be too expensive to be even able to run open source models even for people who can afford it today.
I’m excited for what’s coming. I’m working a lot in silence like a maniac, it’s advancing well, I’m sure the results will follow very soon.
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
KyberSwap Elastic lost ~$48M in 2023 because one tick crossing was skipped.
To get it, you need 1 idea from Uniswap V3 style pools:
Each LP position adds liquidity L only inside its range.
The pool marks the edges:
+L at the lower tick
−L at the upper tick
When a swap moves the price across a tick, the pool applies that marker.
That's how it knows how much liquidity is active right now.
The attack:
1. Add liquidity L in a very narrow range.
2. Swap so the price lands just past the upper tick.
3. A rounding issue in the swap math → the pool does NOT register the crossing → L should be removed, but stays active.
4. Swap back down across the same tick
→ the pool applies the marker again → active liquidity = 2L
Now the pool prices swaps as if it had 2L.
But only L is backed by real tokens.
More liquidity = more tokens out per swap.
The extra came from the other LPs.
Auditor lesson for concentrated liquidity forks:
Check every place the price can end exactly on or just past a tick boundary.
"Price moved past the tick" and "tick was crossed" must always agree.
Learn Harness Engineering 🚀
Found this excellent open-source course by Walking Labs, covering 14 lectures and 8 hands-on projects on building more reliable AI coding agents.
What's inside:
→ Agent instructions (AGENTS.md / CLAUDE.md)
→ Context management and persistent memory
→ Multi-session continuity and state tracking
→ Task boundaries and scope control
→ Testing, verification, and observability
→ Automated agent loops and workflows
→ Multi-agent orchestration and graph engineering
The practical projects walk you through building a complete agent harness, comparing prompt-only execution with structured workflows.
Also includes ready-to-use templates, verification scripts, and a harness-creator skill.
Perfect for developers working with Claude Code, Codex, and other AI coding agents who want to understand the engineering behind reliable agent execution.
Completely free.
Here's the GitHub Repo: https://t.co/hd2KKJiAlo
@thdxr True. Smart people see all the levers and knobs to tweak and to optimise. Selectively dumb people realise they have the power of the sun at the fingertips for just 20/month