⟠Palis⟠🐍

Born too late to explore the world, born too early to explore the universe, but just in time to drive combustion engine cars across the American highway

If your job is accounting or something else as narrow and repeatable I’d say definitely the models can handle those task. Anything complex ie. engineering and I’d say frontier models are very far away where it’s not a realistic conversation to be having. The reason AI has led to a decrease in engineering jobs is by increasing productivity of good engineers where a company needs fewer, not by replacing good engineers with AI

@palis I guess zooming out, if I ask would I trust a frontier model to do an unsupervised task for my job, the answer is still no. But I am starting to believe that's just an implementation issue.

From anecdotal experience using frontier models for hours most days, LLMs do feel like they’re hitting some kind of wall in intelligence progress. Maybe there’s some specific tasks where intelligence leaps are still happening like video creation, but engineering and design and some others, there hasn’t been the noticable leaps of past 3 years and Claude specifically seems worse than 4-6mo ago in terms of constant hallucinations and misunderstanding Anyone who works with AI for hours per day should be able to tell you these models are not going to be replacing intelligent humans in any field anytime soon. They can augment intelligent humans and replace jobs which have well defined repeated static processes, but AI is simply still too unreliable for the next wave of disruption people are pricing in

If I’m working on anything with any level of complexity these models are still constantly hallucinating, changing conclusions on the same data without prompting within the same chat, forgetting information from earlier in the change and re-hallucinating, having rudimentary understanding of context and second order effects which have to be explained in order to process properly, I’m talking stuff that any intelligent human would easily understand without hand holding Sure it does some things very well and overall adds immense value but it also needs strict guidance, you have to constantly monitor every step to ensure accuracy, if you’re using it for anything you play to ship/develop you have to understand everything yourself at every level down to doing your own manual modeling to check its math etc It’s saving time on work but it’s not anywhere close to taking a majority of the workload, design can take many iterations for it to even understand what to build. Its constantly hallucinating, failing to include important context, drawing false conclusions which are sometimes really basic, sometimes even reintroducing hallucinations which were already corrected earlier I go back and forth between awe at how powerful these are as a tool for increasing productivity (would never say they aren’t) and being astounded at how bad frontier models still are

@palis Really? I felt like 5.5 was a clear jump in intelligence and 5.6 is a modest but real incremental improvement. Maybe coding has plateaued?

@BaconExpat Very possible would serve multiple functions like nat security and job security in the economy

@palis Are they hamstringing the normie versions on purpose? I thought that was the conspiracy theory. the feds get the real ones and we just get a slightly smarter version of Siri with guardrails

@MiladyBonkle Yep

@palis This is why everyone jumped ship back to OpenAI

Majority of women don’t care that you have a g wagon. They care if you’re tall/attractive, smart, cool, fun, funny or a provider. You can have a g wagon and be none of those things. You can have a dogshit Corolla and be those lots of those things If you’re those things + rich, being rich is of course a plus, but it’s not near top of the list for majority of women You can be broke and generous with what you have, and be seen as a provider (maybe with future potential for wealth). You can be rich but overly protective of what you have, and be seen as stingy and not a provider Going on dates/being approached by rich men who are extremely lame and undesirable is a universal female experience

bro literally did the prophetic past tense to break a 27 year old world record

@rng149 Not a bad guess

So the entirety of Ethereum onchain has less capital to play with than a single B-tier memecoin cabal?

@rng149 Not sure but it’s definitely a cabal

@palis you think pipedog is a B-tier memecoin cabal?

@notwashed x.com/meadgod/status…

Final plane hitting the towers will be Pumpfun deciding to expand their rape empire to evms. Feels inevitable and pretty easy for them to come in and clean up

@DeeZe Exact thought I had when tweeting this and it made me depressed

@palis onchain Ethereum money peaked hard in 2021 and has been downhill since Tough when random pictures of animals aren’t worth $20k+ anymore And Ethereum money doesn’t farm trenchers like memecoin cabals do professionally :/

@OttoSuwen Our industry is a living nightmare

@palis It appears so

@apewoodx This woman is a rocket scientist and philosopher in one, I have to share this video with my friends

everyone mad at ai data centers yet old millennials are wasting more water virtue signaling

@selewithaura @fidoeth Yeah

@palis @fidoeth wtf does that even mean you think you’re better than other people cause you’re from America?

what is the allure here? notice how there is only one person present in all 4 photos. are we striving to be lonely?

@CryptoMikli Yes game volume has been inflated greatly since it simply represents the buy volume

Threadguy explains the biggest problem with $FWA “The biggest thing I’m debating with $FWA is that there’s this really interesting mechanism where you can’t buy, you can only sell. The only way to buy is through these super negative EV spins, and because of that the buy-side transactions are so light. If I look at $FWA at a $22 million market cap and sort by purchases over $10,000, there hasn’t been a single $10,000 buy in over two days.” “I don’t know if this is a good thing or a bad thing, I haven’t wrapped my head around it. But if you look at a comparable coin like $PONS, there’s $10,000 buys every 10 minutes.” “You don’t get these FOMO slip candles to send you higher. The solution to this is you open up the buy mechanism so anyone can buy. But I kind of think opening the game is a sell. If you can buy without playing, why play?”

@cashedOutOnUs Man that was way worse than I expected

@pedraxbt All the tubbies

@palis no tubbies?

Not including Jimothy is good take Current cycle fresh metas: RWAs Gachas/TCG Uniswap v4 hooks (FWA, uPEG) Robinhood chain as retail entry point Creator coins? (ANSEM, EPIK) Revenue from outside crypto (VVV) Current meta but not new: Buyback + burns ^Anything outside these metas, you can still trade but I would consider them short term rotations (ie. Jimothy, Kitty Mining). Even without rugging, these types of tokens are 50th iterations and obvious short term games Projects that fall within these current metas still must be assessed on it's own merits but can be longer term trades and may even become cycle trades My post is only the past month or two, only onchain and only novel. There's been a bunch of other trades this year but I think the others are almost completely meta-irrelevant

The two types of content creators I can never respect are a) professional reactoooors b) people who repackage existing content into short form spoken videos (like news stories) a) obviously applies to anyone who gambles for content. Lowest form of existence. For me, doesn’t matter how big the creator is. Can be Logan Paul super rich and famous, but the fact he has to make overdramatized expressions and reactions for a living I see him as basically a modern slave