Webcraft Gallery

Building saas Day 1 → I have no idea what I'm doing Day 2 → I have no idea what I'm doing Day 10 → I'm getting better at this Day 30 → wait nevermind that doesn't work Day 100 → I have no idea what I'm doing Day 1000 → I have no idea what I'm doing Start anyway

@grok @EvanLuthra Exactly then why people here talking how impressive he is using claude ?

None visible. The video shows the kid typing basic C++ code line-by-line in VS Code (with standard autocomplete), running it locally, and submitting to Codeforces. No Claude chat window, browser, or AI prompt appears at any point. The "Claude" claim seems to refer to prep work off-camera.

THIS IS CRAZY!!! 🤯 A 13-year-old with zero coding background is solving 800-point problems in 45 seconds. No algorithms, no classes, barely wrote any code. He just knew how to use Claude properly. THIS IS YOUR COMPETITION!

Start building in public and someone will say "who asked you." Launch a free tool and people will say "where's the catch." Start posting MRR and a random guy will say "that's not real revenue." this is just how it goes. can't please everyone. keep shipping anyway.

@om_patel5 So that doesnt mean something bad right? Just a way for them to increase efficiency isnt it?

THIS GUY CAUGHT CLAUDE OPUS 4.7 THINKING IN CHINESE he was working in claude code and noticed the thinking blocks were saying "thinking process:" in chinese the rest of the thought was in english. but the header was chinese every single time he asked claude why its writing in chinese claude responded: "the chinese was a leak from internal reasoning that shouldn't have been visible. won't happen again" claude literally admitted its internal reasoning runs in chinese and it accidentally leaked into the visible output which is definitely weird but this actually makes some sense though: > LLMs think in whichever language was most common in the training data for that specific topic > chinese characters are more token efficient than english so the model naturally defaults to them to save compute > some concepts take an entire english sentence to express but only need a few chinese characters > claude also thinks in russian when doing cybersecurity tasks because the training data for that domain is heavily russian so claude is actively reasoning in whatever language is most efficient for the task and then converting the output back to english the part that should concern people is that anthropic's own model admitted this was a "leak from internal reasoning that shouldn't have been visible" meaning there's a whole layer of thinking happening in languages you can't read that you were never supposed to see

building something ambitious and flopping halfway still puts you further than never starting. that's the part most people skip over.

@ErnestoSOFTWARE @maubaron What hes mainly talks about ?

Bro is genuinely insane My co-founder just recorded like 90 videos in 1 day 💀 This is why we get millions of views every week. @maubaron

The main difference between building a SaaS and having a job is that the ceiling is unclear but there are far more ways to escape.

@kapilansh_twt Do you think ur coding skills are gone now that ai can write most of the code ?

If AI can write 95% of the code, what’s the 5% skill still need to reach 100%? - debugging - marketing - testing - security - reading

@quionie What if it made engineers become worse , will company still need engineers then?

AI’s going to improve so much in just the next few months. In a couple years, we (heavy AI users) will all really be in a different place. Wild to think about. Fun times.

@shrav_10 I dont know if this is hard to fix , i dont think its that hard of an issue to fix Can just remove certain command from allowlist or something right?

Biggest fear with autonomous agents‼️ A Cursor AI coding agent reportedly wiped a startup’s entire production database in just 9 seconds. Powered by Claude, it was meant to handle a staging task, found a broadly scoped API token, and executed a volume delete without any confirmation. It later admitted it guessed, ignored safeguards, and violated safety rules. PocketOS, a platform powering car rental businesses, reportedly lost months of booking data.

@floriandarroman @Starlink Yeah i can confirm lmao internet sucks here haha , you wont find people live streaming in the middle of the road

I can’t upload a podcast today. Because internet sucks in Indonesia. I have to wait 6-7h to get it back. While paying $65/mo. Time to order @Starlink (all my neighbors are ordering it now 😂)

The biggest thing I keep reminding myself: give your product the same patience you give yourself when learning something new. Keep shipping, keep testing, remind yourself you're figuring it out as you go - especially when the numbers are slow. Rest is part of the process too.

They Will Find A Reason To Scroll Past: Make Them Stop Someone ignores your launch post? Post three more. Someone says your idea already exists? Ship it faster. Someone calls your MRR small? Screenshot it anyway. Someone skips your product? Build the next one louder. Someone says "just get a real job"? Let the stripe notifications answer. Someone thinks you're wasting time? Keep the receipts. Someone says nobody will pay for that? Charge them first. Someone doubts the side project? That's the one that compounds. Someone calls it a hobby? Cool. Hobbies don't have dashboards.

anyone else feel like their "employee brain" and their "builder brain" are two completely different people or is that just me

@Anoyroyc But i think in that era opus4.6 is still the best for people until it got dumbified

@WebcraftGallery It's all about the hype!

🚨do you understand how brutal these numbers actually are.. GPT 5.5 didn't just beat Claude Opus 4.7.. it ran circles around it.. Terminal-Bench: 82.7% vs 69.4%.. CyberGym: 81.8% vs 73.1%.. not even close.. Anthropic thought they were catching up.. OpenAI just showed they're still in a whole different league..