things i am wondering lately:
- what is the reason some scaffolds perform better? is it the prompt? the available tools, esp. the search tool? the loop itself?
- what impact has model-specific prompting on the first question? how impactful is this for the frontier?

Twitter

Bitcoin at Block – the Bitcoin project of Block , the company founded by Twitter co-founder Jack Dorsey – has just announced the launch of a free Bitcoin Faucet, marking the return of the Bitcoin distribution model...

Jack Dorsey announces a $1 million project to give away free Bitcoin.

MicroStrategy founder Michael Saylor asserts that Bitcoin has won the global narrative war, specifically naming BIP- […]
The article, "Michael Saylor Warns BIP-110 is the Biggest Threat to Bitcoin; Four-Year Cycle is Dead, Institutional Capital is the Real Protagonist," was first published on BlockTempo, the most influential blockchain ABMedia media.

Michael Saylor warns that BIP-110 is the biggest threat to Bitcoin; the four-year cycle is dead, and institutional capital is the real protagonist.

Global gold and foreign exchange reserves have reached approximately $4 trillion, surpassing foreign holdings of U.S. Treasury bonds ($3.9 trillion) for the first time. This is the largest increase since 1996…
The article, titled "Gold Surpasses US Treasury Bonds for the First Time in 30 Years! Becoming the World's Largest Foreign Exchange Reserve, Why is BlockTempo Bank Asset Allocation Logic Being Rewritten?", was originally published on BlockTempo, the most influential blockchain ABMedia media.

Gold surpasses US Treasury bonds for the first time in 30 years! Becoming the world's largest foreign exchange reserve, why is central bank asset allocation logic being rewritten?

My hot take guess is it's simply the difference in precision of context windows, and that's it