Roy Scott

@CChirchi @noisyb0y1 Except this literally never happened

@noisyb0y1 that's wild, claude out here saving scholarships!!

19-year-old American student spent his entire scholarship $2,299 on an iPad and Mac Mini. His grandma thought he was trying to hack the Pentagon. He just dropped one file into Claude and went to sleep. CLAUDE.md processed four principles. Written after Karpathy - co-founder of OpenAI - publicly broke down how AI tools actually work. In the morning the system closed tasks a junior dev would take three days to finish. Claude did it in 3 hours. Didn't touch code nobody asked it to touch. Just did exactly what it was told. 45,000 developers installed this in a week. Most of them had never heard of the file before. The iPad became his office. The Mac Mini runs while he's in class. An agency does the same thing for $15,000 a month. He pays $20. Four lines of logic. And the question is no longer "can a student compete with an agency" - but whether an agency is needed at all.

@parlettodotbet @noisyb0y1 It’s a lie, though. A straight up lie that has been reposted with the same video and different scenarios. It’s all hype. Stop believing it all

@noisyb0y1 this investment turned out really good for him smart decision

@newlinedotco @noisyb0y1 You are an idiot

the junior dev comparison is actually the most interesting part of this. three days of work in three hours sounds like a scaling win but the real shift is in the evaluation loop. most developers lose those three days because they’re manually verifying every line of generated code. when you use a structured protocol like that you’re moving the verification from human eyes to automated judges which is why the output actually stays within the lines nobody asked it to touch

@imis4n @noisyb0y1 Spoiler: it doesn’t and this is a lie

@noisyb0y1 wait so he just... wrote instructions in a markdown file and claude actually followed them perfectly? that's kind of insane if it works that consistently

@noisyb0y1 🙄

@auOzAI @sickdotdev Yay AI generated fucking useless post! 🎉

@sickdotdev Love this, super practical stack for getting an MVP out fast and cheap. Claude does the heavy lifting, Supabase + Vercel handle infra, and Stripe keeps the lights on. Solid list!

If you want to build a startup: - Claude = coding. ($20/mo) - Supabase = backend. (Free) - Vercel = deploying. (Free) - Namecheap = domain. ($12/yr) - Stripe = payments. (2.9%/transaction) - GitHub = version control. (Free) - Resend = emails. (Free) - Clerk = auth. (Free) - Cloudflare = DNS. (Free) - PostHog = analytics. (Free) - Sentry = error tracking. (Free) - Upstash = Redis. (Free) - Pinecone = vector DB. (Free) Total monthly cost to run a startup: ~$21 Still any excuses?

@sickdotdev Or, hear me out: have a skill set that is valuable in the market. Every jackass is using AI to create a useless startup right now. You will fail

@Jaselton @tonysimons_ K. You’re a fucking moron. We get it

@tonysimons_ Good luck with that, I agree with Alex

This is a wild take -- and I'm a big Alex Finn fan, let me just preface this rant with that. Suggesting people pay $1,000 a month in this economy is ludicrous. I pay for: > ChatGPT Plus ($20) > Claude Pro ($20) > MiniMax Highspeed Token Plan ($40) I'm seriously considering throwing in another $10 for a @NousResearch portal subscription for the immense value it brings to Hermes users. This is more than enough inference for 99% of the people using AI agents right now. Unless you're doing some really deep shit, spending $1,000 a month on API fees is just insane. Remember, this is coming from an influencer with almost half a million followers here and a thriving YouTube channel. $1k a month likely isn't hurting Alex's wallet a bit. But for many of you reading this, it's just flat out not necessary. Especially for smaller businesses who don't have or don't want to spend an extra 1k a month. On a side note, I'm really surprised to see Alex still riding the OpenClaw wave as hard as he is. Hermes is a superior product. He uses it, so he can't deny that fact. Kinda feels a bit irresponsible to be still plugging OpenClaw so hard when there's no argument as to which is a better option for 99.9% of the people watching his videos. Alex, I doubt you'll see this (or even care if you do)... I'll still keep watching your videos, but I'd really love to see you start covering Hermes more, brother. Do your subscribers that favor. The views will still keep rollin' in. (Spoiler alert: building a mission control dashboard doesn't count as doing really deep shit either, by the way. Save those premium tokens and build something useful, folks.) /EndRant 🙃

@tonysimons_ @trentster Human in the loop won’t be phased out for a VERY long time

@trentster Human in the loop will be necessary for the foreseeable future. I predict we get phased out by Christmastime next year.

@manzombie_eth @tonysimons_ Except it’s not. Why? It’s generating unmaintainable slop that a human has to deal with. It’s far more expensive at the end of the day

I don't think Alex is saying everyone should pay $1k. He's saying if your work demands that much compute, it's still cheaper than a human. That's a proportionality argument, not a spending recommendation. You can't hold an influencer responsible for how people misread basic logic.

@tonysimons_ @Josh_is_regen Is he? He’s using the exact same approach as countless AI hype dipshits I see on here

@Josh_is_regen He’s a genius when it comes to driving the hype train.

@tonysimons_ Anyone following Alex Finn is a true moron. He’s simply an AI hype influencer, nothing more. He makes money off the AI hype grift and has zero integrity

@TaskPoolAI @itsolelehmann 🤣🤣🤣 no you aren’t. You’re just running the same overly optimistic, ambitious grift as the rest. Fuck off with your pathetic marketing attempt

@itsolelehmann Biggest vibe shift nobody's talking about yet ... AI agents still can't do anything in the real world We're fixing that 🌊 @TaskPoolAI

if you've been paying attention, you know there's been a serious vibe shift in AI this month 3 major AI labs (anthropic, openai, amazon) all just went hard into biology. they're literally telling you this is where the next wave of ai value capture is here's what happened:

@itsolelehmann Nope. You’re just reading into it and trying to drum up hype as usual

@stevemordue @sukh_saroy Prompt structuring natural language in JSON is simply fucking stupid

@sukh_saroy Tell that to the 99% of vibers who don’t know what JSON is

Prompt engineering is dead. Nobody is writing english to chatbots anymore. The best outputs are coming from people who write prompts like code. It's called json prompting. and once you see it, you can't go back:

@DotNetEric @MakJoris @sukh_saroy Nobody gives a fuck about the vibe.

@MakJoris @sukh_saroy that prompt vibe feels off... but you're onto something there

@Tomas_Zubiri @sukh_saroy Absolutely. What he’s aiming for here is actually more work than structuring a more complete natural language prompt

@sukh_saroy There's more text in their training data than json.

@sukh_saroy This is no more effective than effectively articulating in natural language. You are a dumbass

Now watch what happens when you switch to json: { "task": "write a tweet", "topic": "dopamine detox", "style": "viral", "length": "under 280 characters", "tone": "punchy and contrarian" } Same request. zero ambiguity. the model stops guessing and starts executing.

@sukh_saroy Prompt engineering was never anything, nor is this.

@thealikazai @SethSHowes Then you should read more

@SethSHowes Most interesting read iv read in a while. Thanks.

I sequenced my genome at home, on my kitchen table. I wrote up exactly how I did it - the equipment, protocol, theory, and cost: iwantosequencemygenomeathome.com