lcm_in_pangea's profile picture. ❤💻💵 | CTO/Founder @in_pangea

James

@lcm_in_pangea

❤💻💵 | CTO/Founder @in_pangea

置頂

This is the culmination of a lot of work across our platform, where we can now produce research in a few days from thesis to proof. LLMs + our framework make mincemeat of millions of Aave data points.

Have we reached the end of the bull run? Mega thread incomming. Our DeFi Heat Index 🔥 may offer some hopium! Read our blog post to find out more about this new indicator of market risk and what it suggests about the current market cycle... 👇

in_pangea's tweet image. Have we reached the end of the bull run?

Mega thread incomming.

Our DeFi Heat Index 🔥 may offer some hopium! Read our blog post to find out more about this new indicator of market risk and what it suggests about the current market cycle... 👇


This with a Mac ultra 512 + a number of external gpus could be amazing if we split the heavy compute parts with low bandwidth on the external cards which keeping a large moe model on the main unified memory. Could be the biggest bang for buck for self hosting

Confirmed new @__tinygrad__ on Nvidia GPU (5070) works on macOS Sequoia (M1 MAX). Fun!



James 已轉發

Crossing fingers that this can alleviate my Claude bill

Super pumped that MLX now supports MiniMax-M2! Here’s a quick guide to try it out on your Mac (Apple Silicon). Have fun! 😎 Requirements: 🍏 Apple Silicon Mac (M3 Ultra or later) 🍏 >=256GB of unified memory huggingface.co/MiniMaxAI/Mini…



Anyone managed to run Minimax M2 locally?


You will be subjects of the EU and you will enjoy it

so basically if you are born in a certain geography, that place now has the right to take a cut of all of your money, forever into the future, no matter where you are think about how insane that is for a second I mean like really think about it they literally own you entirely



James 已轉發
mvpatel2000's tweet image.

Today, we announced that we plan to expand our use of Google TPUs, securing approximately one million TPUs and more than a gigawatt of capacity in 2026.



AI can't build great datasets -- without you. Yet.

We wanted something that reflects not just how we think about the world of AI coding agents. But also our role as developers in it.

jayair's tweet image. We wanted something that reflects not just how we think about the world of AI coding agents.

But also our role as developers in it.


You know this is a funny joke, but Metamask and others ought to talk to @in_pangea we've built in mitigations for these long tail risks.

Hi! Sad news ! I just got fired from my job at MetaMask, I was the one responsible for making sure when you first load in it shows a 0$ balance so you get a minor heart attack and think you were wallet drained before it refreshes and shows the real balance a few seconds later.…



Can’t wait to see a Mac ultra with a couple of external gpus.

NVIDIA over USB4 on MacBook is ready to try! * ADT-UT3G dock + any 30/40/50 series GPU * Disable SIP * Install driver `extra/usbgpu/tbgpu` * Install NVK compiler `brew install tinymesa` * Test with: `DEBUG=2 NV_NAK=1 NV=1 python3 test/test_tiny.py TestTiny.test_plus`

__tinygrad__'s tweet image. NVIDIA over USB4 on MacBook is ready to try!

* ADT-UT3G dock + any 30/40/50 series GPU
* Disable SIP
* Install driver `extra/usbgpu/tbgpu`
* Install NVK compiler `brew install tinymesa`
* Test with:
`DEBUG=2 NV_NAK=1 NV=1 python3 test/test_tiny.py TestTiny.test_plus`


Really enjoying @solid_js feels a lot more native than React and running into far fewer issues with highly complex applications. Stores make sense, signals make sense. Just need to get the LLM to understand it a little better :)


Pangea Studio's agentic data in action.


Pretty desk ornaments released

Nvidia DGX 128gb $4000 mini pc benchmarks are in.... Barely gets 11tps on gpt-oss-120b fp4... 6tps on qwen3-32b-fp8 even with sglang optimizations Yeah this is a flop, buy an actual gpu.

nisten's tweet image. Nvidia DGX 128gb $4000 mini pc benchmarks are in.... 
Barely gets 11tps on gpt-oss-120b fp4...   6tps on qwen3-32b-fp8 even with sglang optimizations
Yeah this is a flop, buy an actual gpu.


James 已轉發

After raising pool caps to $50m all @yieldbasis markets filled up in under an hour with some very large deposits. 0x196a2A9A22C2fD8f5107e97Df9ad14A23e81982B added 112 cbBTC and WBTC, with over $31m in liquidity provided by this address alone.

in_pangea's tweet image. After raising pool caps to $50m all @yieldbasis markets filled up in under an hour with some very large deposits.

0x196a2A9A22C2fD8f5107e97Df9ad14A23e81982B added 112 cbBTC and WBTC, with over $31m in liquidity provided by this address alone.

Generated data from Pangea Studio

In a single 4-hour window yesterday, Aave experienced a liquidation cascade: $111M wiped out. 787 positions liquidated.



I get to listen to Joe everyday and I still learnt something from his deep dive into his research supporting the 300M USD decisions @yieldbasis and @CurveFinance.

Leviathan Livestream: Llama Party feat. JoeWait of Pangaea x.com/i/broadcasts/1…



James 已轉發

Refueling is live on @yieldbasis and coming soon to @CurveFinance. But what is it? The new Twocrypto pool implementation automates the adjustment of concentrated liquidity - but rebalancing has a cost. Refueling covers this cost to optimise liquidity depth and lower slippage🧵

in_pangea's tweet image. Refueling is live on @yieldbasis and coming soon to @CurveFinance. But what is it?

The new Twocrypto pool implementation automates the adjustment of concentrated liquidity - but rebalancing has a cost. Refueling covers this cost to optimise liquidity depth and lower slippage🧵

James 已轉發

Due to its relatively high 1% pool fee, @yieldbasis receives low volume when BTC volatility is low. Trades cluster in vertical bands along big price moves. "Market volatility is your yield"

in_pangea's tweet image. Due to its relatively high 1% pool fee, @yieldbasis receives low volume when BTC volatility is low. Trades cluster in vertical bands along big price moves.

"Market volatility is your yield"

James 已轉發

Very interesting article from our friends at @in_pangea . @yieldbasis generating ~8% incremental volume for @CurveFinance . Some thoughts: One key point is that YB has focused on providing extremely deep liqudity in a specific pair (crvUSD and BTC through 3 wrappers). This…

We investigated how much volume YieldBasis generates for Curve. We found: - Every $1 traded on YieldBasis generates at least $1.089 volume on Curve - >88% of this volume was due to crvUSD swaps - YieldBasis has a disproportionate impact in increasing Curve’s market share

in_pangea's tweet image. We investigated how much volume YieldBasis generates for Curve.

We found:
- Every $1 traded on YieldBasis generates at least $1.089 volume on Curve
- >88% of this volume was due to crvUSD swaps
- YieldBasis has a disproportionate impact in increasing Curve’s market share


James 已轉發

We investigated how much volume YieldBasis generates for Curve. We found: - Every $1 traded on YieldBasis generates at least $1.089 volume on Curve - >88% of this volume was due to crvUSD swaps - YieldBasis has a disproportionate impact in increasing Curve’s market share

in_pangea's tweet image. We investigated how much volume YieldBasis generates for Curve.

We found:
- Every $1 traded on YieldBasis generates at least $1.089 volume on Curve
- >88% of this volume was due to crvUSD swaps
- YieldBasis has a disproportionate impact in increasing Curve’s market share

James 已轉發

👀

After lifting pool caps to $10m and BTC sending to ATH @yieldbasis pools are going brrrrrrrr

in_pangea's tweet image. After lifting pool caps to $10m and BTC sending to ATH  @yieldbasis pools are going brrrrrrrr


Loading...

Something went wrong.


Something went wrong.