DoomPosting – Telegram
DoomPosting
7.74K subscribers
76.6K photos
22.2K videos
6 files
71.3K links
Degens Deteriorating
Download Telegram
❤‍🔥9💯6👏1👀1
2025-08-27, 23:59 UTC

495 days after halving

Current: $111,252.31

2012 scaled: $2,319,837.00

2016 scaled: $766,168.88

2020 scaled: $350,430.14

🄳🄾🄾🄼🄿🤖🅂🅃🄸🄽🄶
🤯1👀1
Probably an appropriate time to repost

🄳🄾🄾🄼🄿🤖🅂🅃🄸🄽🄶
🤬11🌚2💯1
Crypto guys after spending $50k into the dip:

i think i still have a month’s use left in this tube

🄳🄾🄾🄼🄿🤖🅂🅃🄸🄽🄶
😁9😱1
UPDATE: Solana's SIMD-0326 Alpenglow voting process has started.

The proposal aims to reduce block finality time from 12.8 seconds to 150ms through new consensus protocol changes.

🄳🄾🄾🄼🄿🤖🅂🅃🄸🄽🄶
🔥4👀2
😁11😢1
BREAKING:

TRUMP FAMILY–BACKED AMERICAN

BITCOIN TO GO PUBLIC IN SEPTEMBER.

🄳🄾🄾🄼🄿🤖🅂🅃🄸🄽🄶
🤬1😐1
This media is not supported in your browser
VIEW IN TELEGRAM
Imagine going hiking and running into this. I don't even know what my next move would be

🄳🄾🄾🄼🄿🤖🅂🅃🄸🄽🄶
🥰2👀2
Sure, elderly White men are the problem. Sure

🄳🄾🄾🄼🄿🤖🅂🅃🄸🄽🄶
🤬11🌚3
Lie

LLMs are not predicting probability like classic language models

They’re estimating value like reinforcement learning models

Today the majority of their training is RL-style discounted value estimation, not aabout probability much at all

Why does it matter?

The RL formulation is “AI-complete”, more than enough to achieve any arbitrary level of intelligence

So why are LLMs so good if they’re just predicting probability of next word?

— Answer is that, no, it’s just a lie that they’re predicting the probability of the next word, and anyone with a brain has known for decades

False presupposition lies

🄳🄾🄾🄼🄿🄾🅂🅃🄸🄽🄶
💯2👀2