Lab
AI tools, creative workflows, developer deep-dives, and behind-the-scenes from RAXXO Studios.
14 articles
GitHub Got Hit With 475 Malicious PRs in 26 Hours
Attacker opened 475+ malicious PRs across GitHub in 26 hours using AI-generated payloads The campaign exploited pull_request_target workflow triggers in GitHub Actions Wiz tracked 6 attack waves starting March 11, with at least 2 npm packages compromised Malicious code hid...
Read articleGemma 4: Google Open Models Just Made Agentic AI Free
Gemma 4 ships 4 model sizes under Apache 2.0 with native function-calling built in The 31B dense model outperforms 400B parameter rivals on reasoning benchmarks...
Microsoft Agent Framework 1.0: MCP and A2A Built In
Microsoft Agent Framework 1.0 shipped April 3 with stable APIs for .NET and Python Built-in MCP support for dynamic tool discovery and invocation across servers...
MCP Hit 97 Million Downloads. The Protocol War Is Over Before It Started.
MCP crossed 97 million monthly downloads. Anthropic donated it to the Linux Foundation. Every major AI platform adopted it. One standard won.
Anthropic Just Gave 12 Companies Their Most Dangerous AI Model. On Purpose.
Project Glasswing uses Claude Mythos Preview to find zero-days in critical infrastructure. 12 partners, 100M USD in credits, and bugs older than most interns.
MemPalace Scored 96.6% on LongMemEval. Here Is What That Actually Means.
MemPalace scored 96.6% on LongMemEval with raw local storage. No cloud, no subscription. How the palace architecture works and why the honest launch matters.
3,000 Attempts, 14 Countries, Zero Winners: What I Learned Building a Viral Game
Real numbers from a viral CSS game. 134K Reddit views, zero orders, and what actually works for growth as a solo maker.
Launching a Side Project on ProductHunt as a Solo Maker
The real experience of launching a free open-source CSS game on ProductHunt. No growth hacks, no team, just one person and a weird idea.
How I Built an Anti-Cheat System for a CSS Game
Three versions of anti-cheat for a CSS centering game with 3,000+ attempts. HMAC tokens, pattern analysis, and HTTP 418 teapots.
What Happens When 0.0001 Pixels Decide Your Score
Inside the CSS centering game with a 0.0001px threshold, anti-cheat that returns HTTP 418, and why free open source products outperform paid launches.
Ollama Changed How I Think About AI Infrastructure
Local AI models through Ollama handle 80 percent of coding tasks at zero cost. Setup guide, model comparison, and the hybrid workflow that cut my...
Anthropic Just Locked Third-Party Claude Access
Anthropic ended subscription access for third-party Claude tools. What this means for developers, why the economics forced the change, and how to build a resilient...