Grigory

@om_patel5 Switching to Chinese wan't enough...

I taught Claude to talk like a caveman to use 75% less tokens. normal claude: ~180 tokens for a web search task caveman claude: ~45 tokens for the same task "I executed the web search tool" = 8 tokens caveman version: "Tool work" = 2 tokens every single grunt swap saves 6-10 tokens. across a FULL task that's 50-100 tokens saved why does it work? caveman claude doesn't explain itself. it does its task first. gives the result. then stops. no "I'd be happy to help you with that." no "Let me search the web for you" no more unnecessary filler words "result. done. me stop." 50-75% burn reduction with usage limits getting tighter every week this might be the most practical hack out there right now

Today I lectured on Reactive Frontend Development with Vue.js. The concepts are fundamental, yet the leverage they give #CS students is enormous. Nearly every product has a #web layer. And until #LLMs consistently handle frontend optimization, that knowledge remains a core skill.

@LunarResearcher Could you please provide with your token usage on that project?

My girlfriend asked why I was smiling at 4 AM. Showed her the terminal. "What are all those green numbers?" $1,129. Made while she slept. "Doing what?" Nothing. Claude scanned 14,000 wallets, found 47 that never lose, built a bot that copies them. She watched for 10 seconds: +$3.87 captured +$6.42 captured +$12.71 captured "It just keeps going?" Every few seconds. New line. New money. "How much did you start with?" $300. Now $1,429. Eleven hours. Asleep. "What does it do?" Buys at $0.48. Sells at $0.52. Pockets $0.04. Who wins doesn't matter. "That's legal?" Citadel does this on NYSE daily. 400 engineers. I have one screen. Copy my bot here: @lunar" target="_blank" rel="nofollow noopener">kreo.app/@lunar She looked at the P&L curve. Never dips. Just climbs. "Can you make me one?" Setting hers up now. She still doesn't get how it works. The bot doesn't care.

So why is there still no system where a student gets not just a pass/fail, but actual iterative feedback with multiple attempts? Has anyone tried building this for real? Or are we stuck in the "one attempt - one grade" paradigm forever?

But nowadays AI can already give feedback on solutions that's not worse than an average instructor. It can explain what went wrong, ask a guiding question, and let you try again. Being a student again makes it clear for me, that this system is kinda outdated.

The more I look at modern learning platforms, the more I feel they're optimized for grading, not for learning. You submit -> you get "wrong" -> you're left guessing what exactly went wrong. No feedback. Just a binary answer. Meh...

P.S. I don't yet have any info on if it's blocked anywhere else, it's just what I see when i show the product to my friends here, on the island.

My working theory is operator-side filtering triggered by EU compliance heuristics (GDPR / ads / tracking signals). If that’s true, this is worrying: ISPs acting as soft compliance enforcers, silently breaking reachability without clear diagnostics.

No malware, no phishing, no shady scripts. The only real red flag: no published privacy policy or terms yet, and no consent framework in place.

Ran into a very odd production issue: my website is reachable for most users, but on one major mobile ISP in Cyprus (EU) it’s partially blocked and flagged as a “dangerous website” for some subscribers.

So it has been a good reminder that some of the hardest problems in product development aren’t even technical sometimes - they’re legal, regulatory, and structural. I'm gonna keep you posted on this topic.

Right now, the pool of truly good options is surprisingly small. I’m still searching, testing, and weighing compromises.

One of the most unexpected obstacles I’ve been dealing with lately has nothing to do with code, infrastructure, or scaling. It’s consent managers. Specifically, trying to find one that works well without compromising the product.