ceappot @ceappot

Quick recap of my Breakpoint talk where I covered - propAMM background - Tech Dive: Reverse Engineering a propAMM - Tech Dive: Tools and Techniques for RE - propAMMs vs. Toxic Flow

keyboard_arrow_left Previous keyboard_arrow_right Next

gm gang ☀️

I’ve got my @xeetdotai Creator Cards to give out. Think of these as 1999 Base Set Charizard cards. So who wants a JYP card? 👉 👈

so good that it was threadguy

Bitcoin Magazine @BitcoinMagazine

2 years ago

BREAKING: 🇺🇸 Donald Trump says, “I think crypto has got a great future. Maybe we will pay off the $35 trillion” with it.

959 3K 17K 1.6M 835

What if the greatest minds of our generation are making digital Pokemon card vending machines?

x.com/i/article/2019…

in progress

The Problem with Slop The two places I spend most of my time, Twitter and code, are increasingly filled with AI-generated rather than human-written content. My instinctual reaction is that this is bad, but a common rebuttal is, “Who cares where it came from, judge the work on its own quality. Besides, AI is smarter than a lot of humans”. The Compiler Analogy Imagine a fictional yet realistic story. Some of the smartest people in the world invent a new Rust compiler. It runs build pipelines 100x faster, turning hours into seconds. And sometimes its generated bytecode delivers runtime speedups too! But there’s a catch: it will never throw an error. If your high-level source code is invalid, it won’t notify you, it’ll just compile incorrect behavior. The bytecode will always compile, always deploy, and always run! But its runtime behavior will sometimes be wrong. Should you use this compiler? Why or why not? Advocates claim the efficiency gains are simply too large to ignore. Just take a look at the compiled bytecode and fix any bugs that pop up. Now instead of slogging through pesky errors, anyone can be a programmer. Opponents say it doesn’t help to move faster in the wrong direction. Bytecode is being generated faster than ever, but it’s intense to peer review. They’re concerned that people are no longer publishing source code for quality concerns, and just committing the compiled bytecode instead. AI Writing The fast buggy compiler story is fake, but a parallel story is playing out across all media - books, articles, and tweets. Ideas = Source Code Media = Compiled Bytecode There’s a transformation process that an idea undergoes to make it ready for mass consumption. You might create a pithy catchphrase, or a visual metaphor, or a lengthy essay to spread your idea. This is hard work! To paraphrase Pascal, “I am sorry to write you such a long letter, I did not have time to write a short one.” AI writing speeds up this work 100x or more. You can input a half-baked raw idea, typos and all, and get an aesthetically compelling writeup defending that position in seconds from an LLM. Ask it to do the same for a contradictory idea and it will instantly comply. This is, in essence, a 100x faster Rust compiler that will never throw any error. Your idea, or source code, can be anything. It will “compile” this into a superficially compelling argument regardless of the validity of the idea. The Judgment DDoS Pre-AI, there was a strong correlation between the fundamental correctness of the work and its superficial aesthetics. In code, it is hard to produce code that compiles and passes tests without understanding the underlying structure. It is likewise difficult to write well about a topic without understanding it. Even the extreme case of an intentionally deceiving author must spend serious effort for good aesthetics. Modern LLMs almost always produce aesthetic output. First-order heuristics no longer filter out bad writing or bad code. Everything is superficially plausible. Why is this bad? Because instead of eliminating errors, we obfuscate them. An incorrect human-written article could be dismissed in five seconds, an AI-written one takes five minutes. Bad human-written code would be caught in a PR, bad AI-written code slips through to bring down prod. Turning every invalid idea into an aesthetic presentation is a DDoS attack on taste and judgment. Human errors are far more detectable than AI errors. There must be a correlation between the aesthetics and the fundamentals. Like GitHub PRs are done in source code instead of bytecode, so should written ideas be reviewed and discussed in their human-generated form. That’s the only way to preserve taste and judgment.

I hope you're successful in life, love, and crypto.

It’s not clocking to you that we’re going higher

This is the final shakeout before 170$ $SOL

birb box opening #2 x.com/i/broadcasts/1…

Open Playtest Day 2 now live. Fixes, updates, and UX improvements have been pushed alongside a (primitive) gitbook for players. Play now: sappy.lol/omnia

Bro is trying to pump our bags by any means possible Respect that 🫡

Jack continues to do the coolest shit

@jackbutcher @jackbutcher

8 months ago

Day two done. $1 reserve auction for the 1/1 goes live at $0.

17 7 220 10K 3

Privacy-first institutional L1 blockchain @digitalasset raised $50M in a Strategic funding round from @BNYglobal, @icapitalnetwork, @Nasdaq, @SPGlobalRatings. crypto-fundraising.info/projects/canto…

Market is back to doing nothing

I expect Base onchain to ramp up significantly leading up to the event on December 17th

Wallet checker is now live on @opensea

Wut @Groowut

8 months ago

Wallet checker is now live on @opensea

195 80 728 64K 19