i learned prompt engineering by dealing with lawyers Your task is to amend and clarify the PPM language around side letters DO NOT modify risk disclosures DO NOT set up a new entity for this Changes to the sub docs are OUT OF SCOPE

won't be just one time won't just apply to billionaires doing it once sets legal precedent making it easier to reapply, expand, and beat down legal challenges (a "wholly new tax" has challenges but amendments to an existing tax have huge leeway)

one of my favorite smart contract experiments was Proof of Weak Hands 3D POWH was a transparent Ponzi scheme. 10% buy and sell tax that got distributed as a dividend to existing holders it was marketed and defined as a Ponzi. everyone knew this when they bought. so it became sort of a gambling game and commentary on the crypto market generally this was the crypto winter of 2018. almost everything was down massive, inflows were low, opportunities were bare. this was one of the few places with real activity. the contract had millions at peak (DeFiLlama says over $10mm) definitely from the 4chan /biz/ sector of crypto. the memes and shilling were insane. that group was heavy into Chainlink which had pumped massively despite being rough around the edges, that community was a good place to be for crypto natives. direct onchain participation, experimentation. early "money legos." precursors to DeFi which was a big part of the runup in 2020 onwards. plus just chaotic fun

an update for my fans. i'm going back to work as a principal software engineer took some time off, found enlightenment, got ripped. promised my wife no new business ventures for a while, so here we are i feel like Rip Van Winkle who went to sleep and woke up in a new world where the old software engineering job is dead so what's my edge now? still more domain expertise than the models in most areas. for now. tick tock my most valuable tool/IP isn't my personal stack anymore. instead it's a context system i built and curated datasets that i feed to agents ive built context harnesses for small, underpowered local models which work well. offline, private, etc. not highly valued (yet?) but feels like token and hardware optimization could/should be valuable i've built intuition around what needs my attention vs what to hand off entirely to agents. if the dev workflow has humans as the bottleneck, then there's still alpha there i manage engineers. which involves managing how they manage their agents while i manage my agents get busy living or get busy dying

massive loopholes and open questions regarding liability and AI agents you can decentralize and/or anonymize the creation of software 2023 crypto + AI bubble you had "fair launches" and truly autonomous agents attributing fault to an actual person might be impossible

you don't hear about OpenClaw anymore because "claws" didn't work as long-lived AI personal assistants i found OpenClaw useful and built specialized agents that i still use daily, but i run them with PicoClaw to save tokens (OpenClaw's token overhead is massive)

got blasted with token spam this week then the coins pump what did they know

the AI dooomer argument is that a future ASI / AGI is too potentially dangerous, almost axiomatically so but current AI is not dangerous, except maybe in cybersecurity, but excels at coding so why not use current AI to build safety and defenses why "pace"

OpenAI is pulling free credits for math undergrads in a math marathon, basically because career mathematicians have hurt egos over recent developments this will happen 100x in science and medicine and all other fields for the exact same reasons

1984 is the real risk of AI Jacob Coxon worked for an NGO called "Newspeak" and now trying to fearmonger us into 1984 in 1984, TVs were two-way screen where men physically surveilled people in their homes this sort of thing could be actually be possible at scale with AI

look your haters dead in the eye and ask them "what's the concern?"

i remember in DeFi summer i told an investor how a position earned 5% in a month by staking he said "obviously that's unsustainable" meanwhile i was in pools with 10,000% APY obviously knew 5% is unsustainable and 10k% is ludicrous but i still funny what a time

golden bull hyperbitcoinization supercycle epochal revanchist zillow dot com we're all going to make it respect the pump is ten million enough to retire

recreating the weekend from first principles by hitting my weekly limits across Claude, Codex, and Kimi on Friday with reset on Monday

"code" is now often called "deterministic code" to distinguish from LLMs or agentic systems went from one syllable to a 6 syllable monstrosity look how they massacred my boy

so Fable solved an obscure math problem (Jacobian conjecture) THE SAME DAY that Anthropic makes you pay extra usage credits for Fable i see you

previous cycles in crypto, technical innovation and optimism created bubbles last cycle was like "let's skip all that and just make a bubble" despite this, we got a fair launch, genuine "everything app" contender (what every fintech is trying to be) Hyperliquid

when cobie said $250k is the easy road he was talking about hype btw

politicians used to just waste your tax dollars now they take your tax dollars and actively try to make your life worse

without looking do you think the total market cap of NFTs is higher or lower than $1b