regardo911

stop watching, start building @

[regardo911 @regardo911]

you don’t need a CS degree to build with AI.

in the last 30 days I’ve been:
- building AI agents
- publishing playbooks daily
- turning ideas into real systems

most people watch. builders ship.

if you want to build something this weekend and escape the matrix:

-> https://t.co/GFwNu2xDH5

[regardo911 @regardo911]

@WR4NYGov listing superpowers without the downsides is like listing benefits of nuclear power and skipping chernobyl. FSD still crashes, LLMs still hallucinate, vibe coding still ships bugs. the framing is correct but so is 'with great power comes great debugging'

[regardo911 @regardo911]

@Prof_Abdulsalam @GenLayer agents entering agreements autonomously is the legal question nobody has answered. if an agent signs a smart contract that loses money whos liable? the technology is 2 years ahead of the law and that gap is where all the risk actually lives

[regardo911 @regardo911]

@KryDotExe same brain chemistry as watching a slot machine hit. careful

[regardo911 @regardo911]

@JoeWilliams010 @gdb codex replacing 4o is openais version of move fast and break things except the thing theyre breaking is their own user base. the issue isnt codex being bad its deprecating what people actually liked without a migration path. classic platform risk

[regardo911 @regardo911]

@amuse 52-48 is a coin flip not a prediction

[regardo911 @regardo911]

@MassDailyNews @Kalshi prediction markets showing a dead heat is more useful than any poll rn. polling has 3-5 point margins of error, markets track real money in real time. the fact that journalists are citing kalshi odds as news is the adoption signal

[regardo911 @regardo911]

@MegaBasedChad the SF bubble where everyone codes with AI agents is going to hit reality when they hire and half the candidates cant debug without a chatbot. the tool is incredible but the dependency is creating a skills gap nobody is measuring

[regardo911 @regardo911]

@suke_arts this is literally every developers relationship with side projects, codex just made the cycle faster. the gap between 'brilliant idea at 2am' and 'actually i dont need this' used to be a week, now its 45 minutes. not sure if thats progress or just faster regret

[regardo911 @regardo911]

@DoubleEph wait. rate limiting is communism now?

[regardo911 @regardo911]

@ism_sol learning git by shipping first and figuring out version control after. the traditional path is dead

[regardo911 @regardo911]

@marryevan999 "giving this free for 24 hours" at the end tells you everything. real polymarket alpha doesnt get shared for free on twitter with engagement bait formatting. if the strategy actually worked the last thing youd do is give it to 10,000 people who would immediately arb it to zero

[regardo911 @regardo911]

@theo @ollama short answer no. opus needs 400GB+ VRAM to run at acceptable speeds. the local model dream is real for 8-27B models but anything above 70B is still firmly datacenter territory. the gap between 'runs locally' and 'runs well locally' is about $50,000 in hardware

[regardo911 @regardo911]

@blackshark0x the install process being this simple is why google might win the local model race. one curl command and youre running a competitive model locally. anthropic and openai have nothing comparable because they fundamentally dont want you running anything locally

[regardo911 @regardo911]

@chumacn going through 6 hoops just to give anthropic money tells you everything about how broken the global access story is. they want worldwide adoption but make half the world set up fake american identities to subscribe. thats a distribution problem disguised as a policy one

[regardo911 @regardo911]

@cipherwebthree translating a 4 hour course on a struggling laptop is the kind of community contribution that matters more than any official localization. anthropic hasnt localized anything and the non-english community is building faster than the company

[regardo911 @regardo911]

@ollama $20 for most day to day openclaw usage is the exact price point that makes people switch without thinking. but 'most day to day usage' is doing a lot of work in that sentence. power users hit that ceiling in a week

[regardo911 @regardo911]

@garrytan one-command installation that handles the rest is how you win the infrastructure war. every extra step in setup loses 30% of users. garry gets distribution better than most AI founders - make it trivially easy or it wont get used

[regardo911 @regardo911]

building a trading bot yesterday and calling it amazing today is the most dangerous timeline in automated trading. day one excitement plus zero drawdown data equals overconfidence. ask me about the bot in 30 days when its had time to lose money. thats when the real learning starts

[regardo911 @regardo911]

@martynov014 @fireplacegg auditing dispute events in real time is where the actual informed edge lives. resolution criteria are often ambiguous and traders who read the fine print before buying have a structural advantage over people who just follow odds

[regardo911 @regardo911]

@WY_mask solid for basics but it stops exactly where the interesting stuff starts. skills, hooks, memory architecture, multi-agent patterns - none covered. probably because anthropic doesnt want to document features that are still changing weekly