Procrypt

it's not enough that I win everybody should win with me

@Debussy100 That people think the only buyer is Saylor

Spent a month mostly off twitter and I've kinda missed you

@thedefivillain @inversebrah not in a direct llm said do xyz way to backtest and implement strategies tho yes

Have you ever used AI to provide you with trade ideas that ended up being really good trades?

@TFTC21 The replies say we have plenty of time to dance on this plateau before we tip off the edge tho

Ken Griffin went home on a Friday "fairly depressed" after watching AI agents at Citadel do work that used to take teams of PhDs in finance months to complete. Done in days. His words: "These are not mid-tier white collar jobs. These are extraordinarily high skilled jobs being automated by agentic AI." This is the head of one of the most successful hedge funds in history saying the people he pays seven figures to analyze markets and structure deals are being replaced by software that works in hours instead of months. Not theoretically. In his own office. Right now. The Coatue deck we covered earlier this week called agents "the biggest unlock" in AI. Griffin just confirmed it from the buy side. The shift from copilots to agents is not a future event. It is already happening at the highest levels of finance.

thx buddy I do try

@theo Turn on remote control and paste it from the app

Just learned it's literally impossible to paste images into Claude Code over SSH. How do you CLI people live like this??

@loomdart Full circle

I am so incredibly bearish on the "global stage platform" version of the internet, and so bullish on lightly gated niche internet microcommunities you can see it best on reddit, some of the giga obscure subreddits that are focused on individual topics are the greatest happiest places on earth, and any popular subreddit is akin to taking a lye bath in a furnace with a toaster

@jeffthedunker @0xGeeGee There are not many things faster than that

@0xGeeGee it’ll be a race to see if MEV searchers can extract gains faster than retards can gamble it away

We saw this playing out (many times) with Synthetix, with the same result of LPs getting rinsed. However, Papertrade try a slightly different approach, so it's going to be interesting to see if they succeed.

@stogolp wherever trabucco went

when they say sell in may and go away, where do we go away to?

@erla221942 ask him if he wants the remaining rare pepes I didn't liquidate for basically zero

@cat_maxxxed it's too hard tho

Engagement on posts really does do better when you don’t drop indiscriminate amounts of F-bombs every three words big sigh

@lurkylearning yes I am

Is everyone asian?

significant phases in your life feel like they were only yesterday, and at the same time like another lifetime lived by somebody else

Trenchers down so incredibly bad

@CryptOrca sha256 last man standing

it's gonna get worse ...

@goodalexander Pretty fun how close good outsourcing and good agent wrangling are to one another eh

Unfortunately this is also good policy for managing autistic forward deployed engineers sigh past lives rearing ugly head

after a couple nightmare sessions using /goal (the Codex agent) -- have some high level heuristics 1. "It's done! (but uhh actually idk what done means). Goal will often build the infrastructure you ask without proper stress tests such that it works for 8 hours and then when you test it, it needs 30 minutes to even test the infrastructure it built So rather than telling Goal what to build, embed the time bound acceptance criteria INTO the goal itself Example: "completion of this goal will entail building 1. a CLI tool 2. a time bound harness to make sure the CLI tool can run quickly 3. manage concurrency and scale tests to multiple [transactions, tickers etc] 4. If batch fails then go and audit code until there is a high % success rate -- i.e use this as a spot test of your functionality 5. Once you are comfortable with CLI performance on single item/ batch move this into a Flask based UX that I can test 2. Embed your ignorance into the prompt. Goal is bad with unknown unknowns. It assumes the user is smart and knows what they want to build (which - let's be honest, is a bad baseline given the fact the user is running /goal to begin with - implies a baseline level of golden retriever mentality) Instruct Goal that "part of this goal is a milestone based set of searches that assume this goal is poorly scoped meaning: a. it is building with the wrong tools b. it has invalid assumptions c. it is building something that already exists that could be copied / or paid for via API d. other failed assumptions After each milestone use OpenAI responses API web queries to construct a series of searches that pressure test execution adversarially. Acceptance criteria of the goal is having run these web searches to pressure test the goal and having implemented best practices" ahhhhhh

@KevinASchofield @tompeck No time for politics at the owl sanctuary, they've got two twits to woo

@tompeck Starmer and Darren Jones off to Norfolk owl sanctuary for the afternoon.

real closure of pear tree productions vibes here...

@kernow_bull np feel free to share in DM, I'm always available to highlight other ppl's best next moves whilst being completely blind to my own

@Procrypt interested in the responses here if you wouldn't mind sharing what you learn, i might be going down the route in the coming twelve months my fear is capital ain't flowing here freely given market dynamics and operational costs

Anyone raised in the UK for any bricks and mortar startups recently? How is the climate?

@Prak_Agrajag @hellojintao i’ll take a 50% winrate all day long

@hellojintao flipping a coin to determine buys and sells

discretionary trader has to be the single most STUPID shit i ever read yeah no one uses any discretion when buying or selling they log in and shit just get bought or sold without any inputs it’s insane no discretion involved