this is one of the best things I've read about AI all week someone spent nearly $5,000 on AI in one month then moved most of the work to Kimi and the bill dropped to $278 so they started trying more things this article explains why cheaper AI can be worth more than just saving money

this might be one of the most useful things i've read about AI agents all week everyone keeps building bigger and bigger agent systems but 17 research papers suggest that more prompts can make models worse, and more agents can actually mean less gets done this article explains what actually makes an AI system reliable

this might be one of the most useful things i've read about AI agents all week everyone keeps building bigger and bigger agent systems but 17 research papers suggest that more prompts can make models worse, and more agents can actually mean less gets done this article explains what actually makes an AI system reliable

this is one of the best things I've read about AI all week and it starts with a 17-year-old trying to run an entire company with 122 AI agents and now the last 40% is breaking his brain because making one AI agent is easy making 122 of them work together is a completely different game this article explains what happens when AI stops being a tool and starts becoming the company

this might be one of the most useful things I've read about AI coding agents all week everyone is building bigger and bigger memory systems but the real problem isn't that agents forget everything it's that every time you switch agents, the next one has to figure out what the previous one already learned this article explains why giving every coding agent the same second brain might be a much better idea

this might be one of the most useful things I've read about AI coding agents all week everyone is building bigger and bigger memory systems but the real problem isn't that agents forget everything it's that every time you switch agents, the next one has to figure out what the previous one already learned this article explains why giving every coding agent the same second brain might be a much better idea

this might be the dumbest way we're wasting money on AI this paper got 97% of GPT-4 quality for just 24% of the cost stop using a Ferrari to go buy groceries your $400/month AI stack probably didn't get the memo

this might be the dumbest way we're wasting money on AI this paper got 97% of GPT-4 quality for just 24% of the cost stop using a Ferrari to go buy groceries your $400/month AI stack probably didn't get the memo

this might be one of the most f*cking useful things I've read about AI agents all week someone built memory into 4 different agent systems and found most of it is complete bullshit the real problem isn't agents forgetting things it's that they remember things they shouldn't trust and once one bad assumption enters the memory, every agent after it can treat it as fact

this might be one of the most f*cking useful things I've read about AI agents all week someone built memory into 4 different agent systems and found most of it is complete bullshit the real problem isn't agents forgetting things it's that they remember things they shouldn't trust and once one bad assumption enters the memory, every agent after it can treat it as fact

this is pure f*cking treasure why one of the biggest assumptions behind modern option pricing might be completely wrong for decades, the math assumed markets move in a nice, predictable pattern this paper shows they don’t in the right hands, this flips the whole game:

this is pure f*cking treasure why one of the biggest assumptions behind modern option pricing might be completely wrong for decades, the math assumed markets move in a nice, predictable pattern this paper shows they don’t in the right hands, this flips the whole game:

this is honestly f*cking insane Grok Bot was released less than a week ago and someone already built an open-source version meanwhile companies are still paying engineers to rebuild the same sh*t from scratch

this might be one of the most f*cking insane things I've read all week there’s a formula that helps price trillions of dollars worth of options it’s been around for more than 50 years now two researchers say they may have found a fatal flaw in it if you want to understand what they found before everyone starts talking about it:

this is honestly f*cking insane Grok Bot was released less than a week ago and someone already built an open-source version meanwhile companies are still paying engineers to rebuild the same sh*t from scratch

this might be one of the most f*cking insane things I've read all week there’s a formula that helps price trillions of dollars worth of options it’s been around for more than 50 years now two researchers say they may have found a fatal flaw in it if you want to understand what they found before everyone starts talking about it:

holy sh*t you can send 300 robots to research something and still end up dumber than before sending 300 AI agents to research stuff is the easy part now, one company called moonshot already built that, it’s fast and it works the hard part is what happens after they all come back 1. picture 300 kids each researching one company, now you have 300 reports and a few minutes to spot which ones are secretly connected 2. ask for a summary and you get every fact but lose how they connect, like reading 300 diary entries and missing that 6 kids wrote about the same bully 3. the fix is stupid simple, don’t ask for a summary, ask for connections, and every connection needs proof, no proof, no connection, period that last rule is the whole trick, an empty list still counts as a real answer, that’s what stops the ai from just making stuff up because two things feel related 100 facts have over 4,000 possible connections between them, that’s just math, and it’s why more agents without this rule means more homework for you, not more answers bookmark this before you forget the “no proof, no connection” rule, and like it if 300 AI agents secretly lying to you sounds as wild to you as it did to me

holy sh*t you can send 300 robots to research something and still end up dumber than before sending 300 AI agents to research stuff is the easy part now, one company called moonshot already built that, it’s fast and it works the hard part is what happens after they all come back 1. picture 300 kids each researching one company, now you have 300 reports and a few minutes to spot which ones are secretly connected 2. ask for a summary and you get every fact but lose how they connect, like reading 300 diary entries and missing that 6 kids wrote about the same bully 3. the fix is stupid simple, don’t ask for a summary, ask for connections, and every connection needs proof, no proof, no connection, period that last rule is the whole trick, an empty list still counts as a real answer, that’s what stops the ai from just making stuff up because two things feel related 100 facts have over 4,000 possible connections between them, that’s just math, and it’s why more agents without this rule means more homework for you, not more answers bookmark this before you forget the “no proof, no connection” rule, and like it if 300 AI agents secretly lying to you sounds as wild to you as it did to me