BLN Craft

@Gutterspawn Looking forward!

Not entirely sure why I wasted a few nights on this, but since I did I may as well share it #pixelart #pixelanimation #Wickedboy #gamedev

New AI coding model drops. My CLAUDE.md, calmly: 'let me reintroduce myself to you too.' 🤝 The model changes. The rules that keep it from hallucinating your codebase don't. #AIcoding #devtools #buildinpublic

@Mahmoud_Zalt the brainstorming phase collapsed faster than anyone predicted. worth watching whether the 43% automating marketing are getting output they'd actually sign off on, or just output they've stopped reviewing.

6 weeks after launching sistava.com, I analyzed data from our first 100 users. No one is brainstorm anymore, they’re all expecting AI to execute the work! What they automate first: - 43% Marketing - 31% Sales - 17% Ops - 9% Support #AI #AIAgent #BuildinPublic #SaaS

@mig4ng @MarcioK the place where this bites hardest is validation logic. let AI generate it, review it, then freeze it — the moment you call the model at runtime to decide if input is valid, you've introduced a bug that only shows up under pressure.

AI is probabilistic, not deterministic. 🧠 @MarcioK People are forgetting that for workflows with clear, definitive answers, traditional code is still king. Use AI to build the algorithm, don't let AI be the algorithm—unless you want your deployments to break. #AI #TechTalk

@semihcelikol @ProductHunt actually useful problem — patch notes are written for lawyers, not players. the real test is whether the summaries hold up when a patch drops that changes three things but breaks fifteen.

🚀 I just launched PatchTLDR on @ProductHunt! Tired of reading 10-page patch notes just to see if your main got nerfed? I built an AI pipeline (.NET 9 + Gemini) to turn massive game updates into 30-sec summaries. 🎮🤖 Would love your support! 👇 #buildinpublic #indiehackers

turns out neural nets are just... numbers doing their thing. 🤓 mind blown? read why your AI model is fundamentally a big matrix multiplication. (no mysticism required) maxleiter.com/blog/weights #ML #DevTools #AI

🚨Watchout guys: 1-Click GitHub Token Stealing via a VSCode Bug blog.ammaraskar.com/github-token-s… #github #hacking #security

CLAUDE.md. .cursorrules. README.md. 3 files that should exist before you write a single line of app code — and what each one does for your AI assistant. dev.to/blncraft/3-fil… #claudecode #cursor #vibecoding

.cursorrules antipattern #1: one massive file for everything. Cursor loads it into every context. Backend rules burn tokens while you fix CSS. Frontend rules clog your API session. Split it. Scope it. 5 more patterns here → dev.to/blncraft/5-cur… #cursor #vibecoding #aitools

Your CLAUDE.md is probably too long, too vague, and structured wrong. 5 patterns that fix it → dev.to/blncraft/5-cla… #claudecode #vibecoding #aitools

Start your next project right. 12 AI-ready starters → blncraft.gumroad.com/l/vibe-coder-k… RT if you know a dev who vibes 🔁

€27. One-time payment. Lifetime updates. From the makers of the CLAUDE.md + Cursor Rules Production Pack (162 rules files, 54 stacks). Pair both for the complete AI-native dev setup. → blncraft.gumroad.com/l/vibe-coder-k… #aitools #vibecoding

We shipped something new. 12 AI-ready project starters for Claude Code & Cursor. Paste one prompt. Ship by Friday. → blncraft.gumroad.com/l/vibe-coder-k… #buildinpublic #vibecoding #claudecode #cursor #devtools

Spent months reviewing AI coding setups across teams. The gap between 'AI is useless' and 'AI 10x'd my speed' comes down to one file: CLAUDE.md. Here's what separates the setups that work from the ones that don't: 🧵

1/ Keep your main CLAUDE.md prelude under 800 tokens. Counter-intuitive but: longer = model ignores your rules. Put the must-follow stuff up top, use @include for framework-specific files. The AI actually reads short files.

2/ Scope rules to file globs. Not "always do X" everywhere — "for *.test.ts: always do Y". 3/ One concern per rule file. auth.mdc, testing.mdc, db.mdc — not one giant rules file. Specific rules get followed. Generic rules get ignored.

Spent months reviewing AI coding setups across teams. The gap between 'AI is useless' and 'AI 10x'd my speed' comes down to one file: CLAUDE.md. Here's what separates the setups that work: 🧵