
contributor: @alphagrowth1 | @AlchemixFi | @DAOhaus
Joined May 2016
- Tweets21.7K
- Following2.6K
- Followers1.7K
- Likes31.1K
Kyle Jacobs | Root retweeted
Pixel art for DOS-AGE's title screen. Hand-sketched, hand-drawn by yours truly.
Kyle Jacobs | Root retweeted
Announcing Fynd InBlock
The first solver run by block builders.
xitter.cf/PropellerSwap/status/2…
Kyle Jacobs | Root retweeted
Run your own router
swap.fynd.xyz?utm_source=x&u…
Kyle Jacobs | Root retweeted
the biggest risk managers and market makers asked us for historical tycho data for 10/10 analysis
and now the data is public
- you can see how sell depth within 1% falls from about $4M to about $0.7M for eth/usdc
- at its lowest onchain spot price was $3,086, while Binance was $3,751
- gas spiked 350x as arbitrageurs tried to capture opportunities
data clearly shows how speed and simulation throughput can give you an edge in such situations
fynd is the only router that can handle these demands without breaking a sweat
dm me if you're building a blazingly fast terminal and need one unified book of onchain liquidity or running liquidations
Go back to 10/10 with Fynd and see how prices moved block by block
fynd-recall.pages.dev/
🧵
Kyle Jacobs | Root retweeted
Ok, spent Sunday trying something else fun.
AVI (Videos) rendered entirely from the smart contract.
Mint a block on @base !
block.graphics/
UI still getting some polish.
I (Claude and ChatGPT) made a fully onchain GIF, seemed like a good use of their time.
ghoulballs.vercel.app/
Kyle Jacobs | Root retweeted
hilarious, someone has setup something to work around the 1 per address mint restriction. I love the blockchain. basescan.org/tx/0x1aa247932f…
I (Claude and ChatGPT) made a fully onchain GIF, seemed like a good use of their time.
ghoulballs.vercel.app/
Kyle Jacobs | Root retweeted
👀 onedollaraudit.com quietly passed 1000 jobs 🎉
Kyle Jacobs | Root retweeted
We were able to fine tune an open weight model (Qwen3.5-9B) on proprietary internal data at @coinbase, and it out performed frontier models, at much lower cost and speed, for a specialized task (fraud prevention in this case).
I think we'll see more of this pattern inside companies. The frontier models are amazing generalists. But increasingly you'll see specialist models out perform on more narrow tasks, either working on teams, or by themselves.
We're making it easy to fine tune models internally for our employees. We even have a skill for it in our internal harness that makes it easy to spin these up and start training them:
"/fine-tune-and-serve-llm I want to fine-tune Qwen for payments risk review. help me get started"
I imagine we'll have some backend services that could be replaced entirely by this new architecture.
Kyle Jacobs | Root retweeted
AI is good at stuff. Attracts capital. Bonds sell off. Unemployment rises. Tax receipts down. Governments can’t pay debts. Bonds sell off more. Massive money printing to pay off debt. New world order.
BREAKING: 🇺🇸 US 30-year bond yield just hit 5.7% for the first time since 2002.
Back then, inflation was only 1.2%, the Fed funds rate was 1.75%, and the stock market was already in a crash, about 40% below its peak.
Today inflation is almost 3 times higher at 3.4%, and stock market is at all time high.
🥪❌ MEV bots can't sandwich you if it's impossible to backrun you.
Every sandwich attack relies on three things:
⒈ Frontrunning your trade, buying before you.
⒉ Pushing the price up with its own trade right before yours executes, so you buy at a worse price.
⒊ The bot selling its tokens back to the pool at that inflated price.
On a standard AMM:
The liquidity pool is symmetrical and bidirectional, meaning the same curve that sells tokens to the front-runner buys them back a moment later, after your trade has pushed the price up.
The bot buys low - pushing the price up, forcing you to buy high - then sells high, keeping the difference.
This is impossible to perform on @CarbonDeFixyz.
👉 Asymmetric liquidity: each strategy has a separate buy curve and sell curve, set by the maker at distinct price ranges.
👉 Unidirectional curves: each curve trades in one direction only. A sell curve sells. It doesn't automatically reverse and buy back.
👉 No profitable sell-back: tokens a bot buys from a maker's sell curve can't be dumped back into that curve at the inflated price.
The only buyer is the maker's buy curve, at the lower price the maker chose.
A bot can still try the front-run, but the back-run doesn't pay. And a sandwich without its second leg is just an expensive trade.
#MEV #sorrynotsorrybots
Kyle Jacobs | Root retweeted
today we're releasing apex-flash-1, our first open-weights model for security research post-trained on real vulnerabilities we found and got paid for.
we are also releasing the abliterated variant for researchers who want fewer refusals in their own authorized workflows.
how we built it: cantina.security/apex-flash
Kyle Jacobs | Root retweeted
Replying to @PrimalGlenn @AskVenice
See this for the average token cost statistics:
xitter.cf/LizThomasStrat/status/…
In my opinion, these two things will be true:
1. The cost for inference will drastically go down in the upcoming years.
2. The average consumer token consumption per day will significantly increase.
Kyle Jacobs | Root retweeted
Anthropic just sounded the alarm on GLM-5.3’s cyber capabilities.
We built on it. OrcaCyber Zero 1.0 is based on the GLM-5.3, post-trained specifically for vulnerability research, exploit reproduction, red teaming, and autonomous cyber workflows.
98.01% pass@1 on CyberGym L1.
1,478 / 1,507 real-world vulnerability tasks.
The cyber model race is here.
And open models are moving fast.
Join private testing: orcarouter.ai/models/orca/or…
































