Agentic AI blueprints and patching treadmill

Two distinct threads dominate today's programming landscape: detailed guides for building production-grade agentic AI systems, and a sobering report on the security costs of AI's rapid release cycle. Meanwhile, Sakana's Fugu offers a novel multi-agent approach, and Deno enters the desktop app space.

Sources

Agentic AI in practice

Two pieces tackle the move from chatbots to autonomous agents, each from a different angle.

#01undercodetesting.comJun 22

From GPUs To Memory: The 9-Step Blueprint For Building Production-Grade Agentic AI Systems + Video -

A detailed 9-step blueprint covers the full stack from GPU infrastructure to UI, aiming to help teams build production-grade agentic systems rather than toy demos.

9 min readRead original

#02sakana.aiJun 22

Sakana Fugu — Multi-agent System as A Model

Sakana's Fugu treats multi-agent coordination as a learned behavior rather than a hand-coded workflow, which could make agent systems more adaptable and efficient.

by Finbarr28 min readRead original

Security and tooling

The rest of the day is about the hidden costs of AI speed and new ways to ship apps.

#03cyberscoop.comJun 22

AI’s constant patching treadmill can be a security problem

Backslash Security's report on Claude Code's patching cadence highlights a growing tension: AI model updates ship so fast that developers can't keep up with security fixes.

#04docs.deno.comJun 22

Desktop apps

Deno Desktop bundles the Deno runtime and a web renderer into self-contained binaries, offering a simpler alternative to Electron or Tauri for building desktop apps with web tech.