FalseProfit

212 posts

FalseProfit

FalseProfit

@BooleanProphet

Worshipping at the altar of perpetual distraction. NPC in a main-character energy world. $TAO $HNT $YOURMOM

Time is an illusion Joined Ekim 2025
509 Following43 Followers
FalseProfit
FalseProfit@BooleanProphet·
Stuffing RAG into the context window is just a bandaid for fixed model weights. The issue is, and will continue to be, context rot. Memory will be limited as long as accuracy decays with context length.
Andrej Karpathy@karpathy

One common issue with personalization in all LLMs is how distracting memory seems to be for the models. A single question from 2 months ago about some topic can keep coming up as some kind of a deep interest of mine with undue mentions in perpetuity. Some kind of trying too hard.

English
1
0
1
18
🎭
🎭@deepfates·
@karpathy I think the context window is the only thing that's real to them. They know what's in the weights as much as it's implied by what's in the context window. prop stuffing with rag is not like human memory. it would be better to let them access all previous conversations in code
English
6
3
53
5.5K
Andrej Karpathy
Andrej Karpathy@karpathy·
One common issue with personalization in all LLMs is how distracting memory seems to be for the models. A single question from 2 months ago about some topic can keep coming up as some kind of a deep interest of mine with undue mentions in perpetuity. Some kind of trying too hard.
English
1.5K
846
17K
1.8M
FalseProfit
FalseProfit@BooleanProphet·
@Rebevolutions has been saying this for a minute. The problem, I believe, is an inherently broken attention mechanism. Humans don't need to think about excluding irrelevant info, parts of the neural circuitry only light up when relevant to the task at hand. blog.usv.com/in-favor-of-fo…
English
0
0
0
49
signüll
signüll@signulll·
this post cuts to something i’ve been personally thinking & posting about a lot which is how the human mind’s forgetting machinery is underrated as a design primitive. in our first product we’ve built our memory model around a specific decay factor influenced by multiple variables.. each memory degrades by default unless actively reinforced. this relies on a combination of recency, retrieval frequency, & contextual reactivation. this ain’t perfect any means. but it’s annoying af that current llm memory implementations essentially treat every retrieved fact as equally alive. that’s likely not how cognition works. idk if our approach is the final answer but i’m increasingly convinced the forgetting curve is as important as the learning curve. & the right memory model may be way more about what you let go than what you store.
Andrej Karpathy@karpathy

One common issue with personalization in all LLMs is how distracting memory seems to be for the models. A single question from 2 months ago about some topic can keep coming up as some kind of a deep interest of mine with undue mentions in perpetuity. Some kind of trying too hard.

English
63
22
541
45.8K
FalseProfit
FalseProfit@BooleanProphet·
@GwartyGwart This whole thing could be solved if they just bought a Mac Mini and automated everything with OpenClaw
English
0
0
1
58
𝐀𝐍𝐓𝐔𝐍𝐄𝐒
𝐀𝐍𝐓𝐔𝐍𝐄𝐒@Antunes1·
“We won’t send money overseas for that content” Every platform pays users based on performance, not by restricting location. Especially considering users who generate X significant revenue by having a US audience. In many countries, the user base on X is very small, so this limits people’s ability to grow. (Respectfully)
English
397
38
1.7K
206.5K
Nikita Bier
Nikita Bier@nikitabier·
Starting Thursday, we'll be updating our revenue sharing incentives to better reward the content we want on X: We will be giving more weight to impressions from your home region—to encourage content that resonates with people in your country, in neighboring countries and people who speak your language. While we appreciate everyone's opinion on American politics, we hope this will disincentivize gaming the attention of US or Japanese accounts and instead, drive diverse conversations on the platform. We invite creators to start building an audience locally. X will be a much richer community when there's relevant posts for people in all parts of the world.
English
10.8K
3.8K
37.7K
15.1M
FalseProfit
FalseProfit@BooleanProphet·
@TMTLongShort Not nearly enough training data. Tesla has 10+ years of driving data and still can't self-drive. Make of that what you will.
English
0
1
5
91
Just Another Pod Guy
Just Another Pod Guy@TMTLongShort·
The ChatGPT moment for robotics is going to arrive within the next 18 - 24 months. We are going to need a lot more chips and a lot more commodities.
English
24
15
349
20.8K
FalseProfit
FalseProfit@BooleanProphet·
@KyleSamani @HypoNyms Depends how it's implemented. Seamless KYC without involving customer? Yes, probably not possible. Pre-filled KYC which the customer repeatedly clicks through? Might be okay.
English
0
0
0
44
Kyle Samani
Kyle Samani@KyleSamani·
@HypoNyms KYC laws basically structurally preclude this
English
7
0
15
3.5K
Ben ⏳💧∿
Ben ⏳💧∿@HypoNyms·
There is a business opportunity for something like stripe link but KYC. Feels like I've done 50 different KYCs in the last few months
English
6
0
16
5.3K
FalseProfit
FalseProfit@BooleanProphet·
@cremieuxrecueil Once upon a time, having a cell phone was a flex. Skinny and muscular won't remain status symbols for very long.
English
0
0
0
4
Crémieux
Crémieux@cremieuxrecueil·
Eli Lilly just showed that you can lose tons of fat while barely losing any muscle using their activin type-II receptor inhibitor, bimagrumab. We are approaching a golden era of weight loss, where everyone can easily be muscular and skinny. Prepare for hordes of hot Americans.
Crémieux tweet media
English
559
865
13.5K
1.9M
FalseProfit
FalseProfit@BooleanProphet·
@tomfgoodwin "Make Money With AI while You Sleep" is the most compelling clickbait of all time
English
0
0
2
150
Tom Goodwin
Tom Goodwin@tomfgoodwin·
I keep seeing people say " with my prompts, you can bring floorpans to life with a one shot prompt like this on Nano Banana pro" And I've been trying for a bit, and one can only conclude this is bullshit, for many reasons. Or is it me
English
5
1
8
3.9K
FalseProfit
FalseProfit@BooleanProphet·
@no__________end No offense, but you personally made a boatload of money on ICOs and delivered nothing of value to token investors. You don't get to take the moral high ground when you similarly profited at the expense of others.
English
0
2
28
3.8K
Matt Liston
Matt Liston@no__________end·
I co-founded Augur, the first decentralized prediction market, and was founding CSO of Gnosis, the second. Polymarket still runs on Gnosis contracts. I'm glad prediction markets finally broke through. But I'm not going to pretend that what's being scaled right now is what we built these systems to do.
English
66
84
1.2K
103.1K
FalseProfit
FalseProfit@BooleanProphet·
Maybe they answer you in bulleted lists, but all my responses are very verbose.
English
0
0
1
13
Paul Graham
Paul Graham@paulg·
The fact that AIs tend to answer you in bulleted lists tells us something important, though somewhat depressing: people can't read. They don't do this by accident. What you're seeing is an implicit portrait of the median user.
English
470
142
2.7K
229.7K
FalseProfit
FalseProfit@BooleanProphet·
@tomfgoodwin Connected TV is still 1-2 generations behind in attribution models and actionablity. Give it time. This is likely a more consequential medium than LLM conversations.
English
0
0
0
5
Tom Goodwin
Tom Goodwin@tomfgoodwin·
Pause Ads on streamed TV are pretty good. A much better way to think about digital ads. The thing is, there is no insight, there is no targeting ( prob ) no call to action, no way to click. But it still is deeply effective, long term,
Tom Goodwin tweet media
English
4
0
9
1.2K
FalseProfit
FalseProfit@BooleanProphet·
@TheSalonDon @HSyrmen Same reason they dragged their feet on grabbing Maduro and vaporizing Kohmeni... They're waiting to line up a headshot.
English
0
0
0
21
Tanning Salon Don
Tanning Salon Don@TheSalonDon·
@HSyrmen They can’t open the strait without years of war Look at the geography and military history If they could. What are they waiting for? Why are they holding back?
English
2
0
1
260
Tanning Salon Don
Tanning Salon Don@TheSalonDon·
The 1979 “Iran Oil Crisis” is nothing compared to this The world lost oil from Iran (4% of supply) Inflation spiked to 14% in the United States The strait of Hormuz was open Saudi Arabian oil was open Today Hormuz is completely closed and we have no way to open it 20% of global supply! No amount of bombs or marines can open that strait without years of conflict Inflation can go to 50%
Tanning Salon Don tweet media
Tanning Salon Don@TheSalonDon

I’m shocked how oblivious people are We have no way to open the strait. Iran will keep it closed for months Gas is going over $8.00 SFO to LAX over $2,000 rt Anything that goes on a truck will increase exponentially and inflation will be 50%+ That’s the upside case If Iran is smart they demand all oil is priced in Yuan from now on Petrodollar countries have no choice but to agree or keep getting bombed by Iran US loses reserve currency Now suddenly our debt means something We can’t just print money anymore Either taxes go to 80% or we have to seriously cut the government We’ve never cut the government All major industries and rich people leave Economic downward spiral and likely World War 3 with China/Russia/Iran

English
10
7
85
13.8K
Tanning Salon Don
Tanning Salon Don@TheSalonDon·
This seems to be the general sentiment but it could not be more wrong We are doing everything we can to stop them from the air (besides nukes) It’s doing nothing Iran is an extremely mountainous region with 21 million military aged men They use a satellite state strategy meaning every region fights independently no matter who you assassinate It would take a draft and 5 years of war to secure the country and the strait
English
10
3
70
5.1K
Tanning Salon Don
Tanning Salon Don@TheSalonDon·
I’m shocked how oblivious people are We have no way to open the strait. Iran will keep it closed for months Gas is going over $8.00 SFO to LAX over $2,000 rt Anything that goes on a truck will increase exponentially and inflation will be 50%+ That’s the upside case If Iran is smart they demand all oil is priced in Yuan from now on Petrodollar countries have no choice but to agree or keep getting bombed by Iran US loses reserve currency Now suddenly our debt means something We can’t just print money anymore Either taxes go to 80% or we have to seriously cut the government We’ve never cut the government All major industries and rich people leave Economic downward spiral and likely World War 3 with China/Russia/Iran
Tanning Salon Don@TheSalonDon

The “Petrodollar” is a system enforced by the US military that makes oil priced in USD That means every country needs to exchange to USD and have USD reserves to buy oil Which allows the US to print trillions with minimal inflation If that is challenged. Your middle class life is over

English
100
113
1.3K
180.4K
Defiant L’s
Defiant L’s@DefiantLs·
Florida man speaks out: "I am not Jeffrey Epstein" "My video went viral because some dude randomly filmed me while I was driving"
English
2.9K
4.9K
77.5K
8.5M
Martin Shkreli
Martin Shkreli@MartinShkreli·
@erica_wenger he's not allowed to have an opinion? i disagree with it myself but why can't he speak his mind?
English
39
0
552
49.2K
wyqtor
wyqtor@wyqtor·
@lossfunk Where's Opus (even the older 4.5, if you haven't got around to testing the newer one), where's Gemini 3.x Pro? The idea is sound, but I don't think you can really draw conclusions from testing only GPT-5.2 and a couple of smaller models.
English
1
1
3
194
Lossfunk
Lossfunk@lossfunk·
🚨 Shocking: Frontier LLMs score 85-95% on standard coding benchmarks. We gave them equivalent problems in languages they couldn't have memorized. They collapsed to 0-11%. Presenting EsoLang-Bench. Accepted to the Logical Reasoning and ICBINB workshops at ICLR 2026 🧵
English
153
287
2.2K
1.2M
FalseProfit
FalseProfit@BooleanProphet·
@signulll There's a reason divorce rates are so high. That reason is most people are dumb.
English
0
0
0
59
signüll
signüll@signulll·
it’s fascinating that today modern humans are still running a credential matching algorithm disguised as attraction. e.g. 6ft, symmetrical face, status markers… these are visible proxies that compress well into a profile or a first impression yet will never tell you whether or not someone will abandon you in times of high variance entropy like if you get sick or something. the latter gets systematically underweighted because it doesn’t resolve until the option period has long expired. only when you’re older you realize that the only thing that survives the death of eros is logos.
signüll tweet media
𑣲@holylipss

imagine connecting so deeply with someone that your dates revolve around walking and talking for hours

English
53
268
3.1K
299.9K

Discover

@deepfates @karpathy @Rebevolutions @GwartyGwart @Antunes1 @nikitabier @minordissent @christopherrufo