brn-maker

I asked myself: if I'm shipping with AI agents every day, why can't I see what I actually built this week? So I built a command center for it. No React. No Node. No cloud. Just open index.html. Here's what's inside 🧵

@orangie Bullish on AI usage but skeptical on these tokens long-term. Remember the PreStocks $OPENAI on Solana? Dropped hard after OpenAI warned about unauthorized transfers. Real share backing helps, but custody/transfer risks remain. What's your entry thesis @orangie

You know how bullish I am on AI, I use it every day Tessera just ended their sale & now they have T-OpenAI available now for trading, now we will have another option to participate in sam altman exposure 😋

@DavidOndrej1 Yet data shows degree holders earn significantly more lifetime. The fix isn't burning college—it's hybrid: use AI to 10x your learning speed while getting the credential as a filter. Self-taught exist, but most employers still use the degree as a screen. Adapt or get screened out

college is a joke

@lukebelmar Attachment leads to suffering. The things you own end up owning you

Suffering is a condition that comes from attaching yourself to things and outcomes. If you desire nothing then nothing can control your mental peace. True freedom doesn't come from having things but from not desiring things. God will take care of you, you can rest in Him.

@DavidOndrej1 Opened my own small team nearby last year — the execution speed here is crazy. Thread on why more founders should look East?

I just opened an office in Katowice most people have never heard of this city it's about to become the next Dubai

Master these = go from fighting Claude to having an actual elite coding teammate. Which one are you most excited to try first? Drop it below Bookmark this thread + watch the full video. Tag a dev who needs this! #ClaudeCode #AI #Coding #BuildInPublic

8. Plugins — Team & third-party extensions. 9. Context Management — Stop blowing up your token limit. 10. Slash Commands — Your custom shortcuts. 11. Compaction — Clean up long chats. 12. Subagents — Spawn parallel brains for big projects.

If you're just starting with Claude Code, these 12 concepts will save you dozens of hours of frustration and failed experiments. This video breaks them down perfectly for beginners. Watch it before you waste time guessing!

Drop a like if you’re watching this weekend. What’s one thing you want to understand better about LLMs? (tokenization, RLHF, data, etc.) RT to help others level up. Follow for more high-signal AI breakdowns.

Pro tip: Watch with chapters on. Pause and replay the data pipeline & RLHF parts. Then try implementing one concept immediately (even a tiny script). That’s how it sticks. The gap between watchers and doers is massive.

This 2-hour Stanford lecture by Yann Dubois just dropped the entire playbook on how LLMs like ChatGPT & Claude are actually built. Everything, Most “AI experts” haven’t seen this level of clarity. Bookmark it. Watch it today.

@killix Context saturation with its own failed attempts is the silent killing The model doesn't just forget, it starts treating its earlier wrong paths as "proven" ground truth and doubles down.

@brn_maker Context filling up is the wall, and it fills with the loop's own failed attempts before cost ever bites. The model keeps trusting its earlier wrong guesses, so it can't climb out alone. The human reset works because someone outside the loop picks which step actually ships.

Everyone's panicking about "AI loops" – models training on their own output, recursive agents, self-improvement spirals. The doomers say it's model collapse or FOOM. They're wrong. AI loops are the most overhyped nothingburger in tech right now.

Bottom line: AI loops aren't runaway superintelligence. They're iterative refinement with better statistics. Stop fearing the mirror. Start using it.The sky isn't falling. It's just getting a better autocorrect. What am I missing?

The real danger? Regulators and safetyists trying to "break the loops" with heavy-handed rules. That creates brittle, lobotomized models. Let them iterate. Messy feedback is how intelligence grows.