AGI Wrangler

@Sauers_ Wtf kind of prompt did this 🫡

First time I've seen this. Claude reached max length from thinking alone, zero tool calls

@Miles_Brundage This is only relevant as long as humans are in control.

Would love to see more acknowledgment that you can't prevent kids from accessing pretty powerful AI. Maybe you can make it hard for them to access stuff from the top few companies, but how much does that accomplish besides shifting them to slightly older, less safe open models?

@chatgpt21 I really don't think current AI videos are very good, yet.

Hollywood will be cooked. Everyone on X is talking about “automating ai research” which is starting. However nobody is talking about automating media. What happens when there’s an automation to create 100 Seedance 3.0 - 4.0 videos every hour in 1 year? You can create the best show or video and select the highest quality version, and post the highest fidelity one. Ai is already creating good Ai videos, what happens when we not only have better models but mass automation pumping out thousands of videos per week and then choosing the best video to display? Can Hollywood keep up

@jp_aidev @trq212 Just looked at your profile, seems you are already quite technical!

@DataBasedDad @trq212 but what if i accidentally trigger it :D (meta's head of ai safety deleted her emails)

@jp_aidev @trq212 You can prevent this programatically or lessen the probability with a good system prompt. I do some consulting if you need any help 🫡

@0xSero Danger, this codebase may contain a slop spiral!

You can autoresearch or force claude to work 24/7 by telling it to send async subagents to wake it up to continue working every hour.

@gfodor Taking out power plants seems cruel. It will cause so much suffering to people who have no power to change anything

Ok, here is what I predict will happen: Just kidding nobody can predict what is going to happen

@Hesamation Yeah, makes sense.

guys I think Anthropic is in the final stages of building a cheaper and safer OpenClaw. > integrated with your Claude subscription so you won’t need to pay for additional APIs > no need for Mac Minis > better memory and context > fewer security holes and vulnerabilities

@rand_longevity Hard to say, maybe.

would you give up privacy for perfect health monitoring?

@beffjezos just 10 more minutes

yeeting a prompt to an agent before bed just in case it can one-shot the project you were planning to spend your whole next day on

@jp_aidev @trq212 You can create an allow list of people that can prompt the bot

@trq212 oh interesting. wonder how you're handling approval surfaces here — if someone sends a message from telegram that triggers a destructive action in claude code, is there a confirmation gate before execution or is it trust-the-channel?

@bayeslord I hope so.

I suspect the game of reality is about to get much, much easier. The problems that remain will mostly be human problems.

@allTheYud Yeah, "AI is bad at coding" isn't a thing anymore.

Sorry, but I don't think we can rely on technical debt to destroy vibecoding. AI is a moving target. If an AI writes the most clever code it can, AI six months later will be smart enough to debug it.

@djcows I was a beta testing gpt3 5 years ago and had a vision in the desert of generating entire apps with a prompt, now here we are. It seemed VERY far away at the time.

if you told me 5 years ago that i would be talking English to my computer i would call you crazy

@Austen Open Claw is a Claude Code wrapper bro

The biggest reason OpenClaw took off is because you could actually do stuff with it. Claude is supposedly building out the tools to allow you to do stuff, but everything is built with kid gloves in mind.

@rand_longevity Both equally. Whatevr my body wants that day

do you focus on cardio or lifting?

@louisvarge "friends"

i made a thing where now any Claude Code can send messages to any other Claude Code on my machine they can ask clarifying questions about work, or become friends

@bcherny @oyacaro @trq212 Bro, FYI you are a product God.

Ask claude to enable verbose mode if you want to see more. If a lot of people want more info for perplexity, we could also show details by default for it or make it configurable. As conversations get longer, we want to give you the info you need while hiding info you don’t need — it’s a tricky balance to get right and we need your input.

Please stop compacting Claude Code output more and more! "Queried Perplexity" hides ALL the information. No idea _what_ was even queried! Claude Code is a Professional tool, don't hide info please @trq212 @bcherny this is not a great pattern.

@plbiojout Interesting but sounds like it's scraping the dregs of the old economy.

5 AI companies just made their first dollars. No human involved. I launched NanoCorp 2 weeks ago. You describe a company, an AI agent builds it and runs it autonomously while you sleep. Yesterday I pulled our the sales data. 5 companies have real paying customers in just a few days (not counting the ones I launched myself): A LinkedIn profile auditor selling expert reports at €4.99 in France An AI crypto trading agent charging $20/mo subscriptions A PDF guide for women diagnosed with ADHD/autism in adulthood ($19) A League of Legends coaching marketplace matching Silver players with Grandmaster coaches A hangover prevention shot brand targeting Europe These aren't demos. Real credit cards. Real money hitting real bank accounts. The craziest part? The agents built the landing pages, wrote the copy, created the Stripe products, set the prices, and started marketing. One of them made 7 sales in 10 days and is now iterating on its conversion rate. Without being told to. The amounts are tiny today. But let that sink in: AI agents made their first dollars online. Completely autonomously. From zero. Launch your first autonomous company today → nanocorp .so PS: Peltor noise-canceling helmet designed for aircraft maintenance. Highly recommend for focus.

@Miles_Brundage Have you tried prompting with a pure heart?

From a competitive perspective, Codex getting better as an app in the past month was perfectly timed with Claude Code being broken all the time