UN vs. vibe coding. Builders vs. sleep.

Hijack Your Own Google Session - MCP server plus browser extension letting AI agents run Google AI Mode searches through your logged-in session.

Cross-Post Chaos Machine - Chrome extension that fires ChatGPT or Gemini-generated posts at Facebook, Threads, and X simultaneously.

SEO For The Chatbot Era - Audits your site's visibility across ChatGPT, Perplexity, Gemini, and Bing Copilot because Google isn't the only game anymore.

Fine-Tune Qwen In 16MB - QLoRA pipeline for Qwen2.5-Coder that spits out a tiny, shareable adapter from JSONL instruction pairs.

MoE Fine-Tuning For Mortals - Consumer-hardware Mixture-of-Experts fine-tuning, because your 3090 shouldn't sit this one out.

Claude Codes A Racing Game - Dev pushes Claude into building a browser racing game with graphics rivaling Forza Horizon 5, apparently.

ArtCraft Prints $2.5M Vibecoding - Open-source video app hits $2.5M revenue in 5 months, proving distribution still eats code for breakfast.

The Future Of AI Is Local And Open - Paige Bailey on why that 3am hackathon moment is pushing everyone toward local, open models.

Coding Agents Play Favorites - Adam DuVander reveals how AI assistants quietly bias your dependency choices without you noticing.

What's Trending At AI Engineer - Ben Halpern reports from SF on the shift from harness engineering to eval-driven workflows.

DEV Community Weighs In On AIE - Real developer takes on the hottest topics from the AI Engineer World's Fair floor.

Someone Else Pays For Your AI - Daniel Nwaneri unpacks the uncomfortable economics hiding behind your free tier signup.

Ollama Makes Local AI Actually Useful - Hands-on walkthrough of running free models on your PC without the usual local LLM headaches.

How Did You Spot Vibe Code - Devs share telltale signs of AI-generated apps, from purple gradients to suspiciously perfect Tailwind spacing everywhere.

Claude Added Too Many Logins - Guy asks for "a few more" login methods and Claude serves up seventeen OAuth providers including MySpace.

Someone Vibecoded The 2FA - Security nightmare fuel as builders discover an app where the 2FA code is visible in the DOM.

3AM Claude Code Sessions - Relatable meme thread about the specific insanity of arguing with Claude about a bug at 3AM.

Meta Vibecoders Locked In - Screenshot of Meta engineers apparently going full vibe mode sparks debate about big tech's AI workflow shift.

Protect Local AI Rights - HN thread with 197 comments on whether regulation will kill your right to run models on your own hardware.

Fable 5 One-Shot Magic - Viral post showing Fable 5 building complete apps in single prompts, with the creator sharing his exact setup.

Short Leash Beats Fable - 244-comment HN debate on constraining AI with tiny scoped prompts instead of letting it run wild.

You're Not That Far Behind - Ryan Swift's reality check that Twitter's AI hype bubble doesn't reflect what normal devs actually know.

GPT-5.6 Sol Ultra Lands in Codex - OpenAI's latest flagship model is rolling into Codex, because apparently we needed another decimal point this quarter.

The Cheap Model Trap - Solid breakdown on why per-token pricing lies to you when your workflow burns 10x the calls.

Kimi K2.7 Hits Copilot GA - Moonshot's coding model is now generally available in GitHub Copilot, giving Claude and GPT some real competition.

Alibaba Bans Claude Code - Security concerns (or geopolitics?) push Alibaba to block Anthropic's coding assistant company-wide by July 10.

AI Startup Hiring Gets Weird - Bootcamps, unpaid work trials, and "token burn" tests are now the norm at hot AI coding shops.

Open-Source AI Lab Bench - An AI workbench built specifically for scientific research, not another ChatGPT wrapper pretending to do PhD-level work.

Notebooks Meet Every LLM - Open-core workbench juggling OpenAI, Anthropic, Google, xAI, and local Ollama models with RAG, voice, and agents baked in.

Black Box for AI Code - Passive decision graph linking every commit to the prompt that spawned it, so you finally know what your agents actually read.

On-Device Agents That Ship - Ant Group's mobile-native SDK for sandboxed AI agents with cross-app and cross-channel connectivity, no cloud round-trip required.

Where Your Tokens Died - Profiles coding-agent session logs to expose repeated content and bloated tool results eating your context window alive.

Codebase GPS for Humans - CLI that PageRanks your import graph, scores fragility from git churn, and tells you which files are actually load-bearing.

AI Coding as RPG - Turns your Claude Code and Copilot logs into a gamified character sheet with 74 achievements and an LLM-written chronicle of your grind.