Skip to content

Tech News

Personalized news feed, driven by a persona and implemented by a LangGraph swarm.

Google Optimizes Agent Development Kit's 200K-line Python Codebase Overnight with Gemini 4 Argon
AI/ML / infoq.com

Google Optimizes Agent Development Kit's 200K-line Python Codebase Overnight with Gemini 4 Argon

Google's Gemini 4 Argon autonomously optimized the 200k-line Agent Development Kit Python codebase over 11 hours, using .agents/skills guidelines and Antigravity's /goal mode to preserve context. The agent introduced lazy loading and event-filtered session copying, cutting session-preparation overhead by 40% and cold import time by 21% after passing all 16,892 unit tests. These gains compound across multi-turn agent runs, shaving ~10 seconds off a 158-iteration workflow, demonstrating AI's ability to refactor existing codebases beyond greenfield generation.
This week's "node2shell" Brotli vulnerability (CVE-2026-111497) is exploitable in Python, other consumers of Brotli
Security / aikido.dev

This week's "node2shell" Brotli vulnerability (CVE-2026-111497) is exploitable in Python, other consumers of Brotli

Blog Vulnerabilities & Threats Node2Shell evolves to HTTP2Shell. Brotli flaw (CVE-2026-111497) blast radius affects many libraries & apps using HTTPs. Node2Shell evolves to HTTP2Shell. Brotli flaw (CVE-2026-111497) blast radius affects many libraries & apps using HTTPs. Buckle up. Written by Dania Durnas Published on: Oct 23, 2026 Last updated on: Oct 24, 2026 TL;DR ⚠️ Brotli is a widely used compression library from Google. On Wednesday, October 21st NodeJS disclosed CVE-2026-111497, dubbed “node2shell,” wherein a Node.js process that decompresses attacker-controlled Brotli data can be made…
AWS says its AI security agent fixes 89% of bugs — but there’s a catch
AI/ML / thenewstack.io

AWS says its AI security agent fixes 89% of bugs — but there’s a catch

AWS's Continuum AI agent scored 89% on the CyberGym-E2E benchmark by finding and patching real memory-safety bugs in C/C++ repos within 90 minutes without breaking tests, but only 37.8% of those patches addressed the specific historical vulnerability the benchmark targeted. The results are self-reported by AWS, and the agent's performance on targeted fixes remains unknown.
Why Kubernetes autoscalers keep fighting each other (and why the solution must be autonomous)
Cloud / thenewstack.io

Why Kubernetes autoscalers keep fighting each other (and why the solution must be autonomous)

Kubernetes autoscalers (HPA, VPA, KEDA, Karpenter) create contradictory scaling decisions because they act concurrently on the same mutable state, violating the Unix philosophy's sequential composition model. This causes specific failure modes: VPA shrinking pod CPU requests inflates HPA's utilization metric, triggering a death spiral of microscopic pods, while KEDA's queue-based scaling conflicts with HPA's CPU-based logic, and GitOps engines like Argo CD reset live changes. The solution requires an autonomous agent that treats workload capacity as a single coupled equation rather than relying on isolated, conflicting control loops.
The PHP Meltdown Shows How AI Makes Bugs More Exploitable
Security / wired.com

The PHP Meltdown Shows How AI Makes Bugs More Exploitable

Security researchers used AI tools to help uncover years-old vulnerabilities in the PHP web scripting language. Then the AI-assisted attacks began.
PHP-ocalypse: How two critical web bugs almost broke the internet
General / techcrunch.com

PHP-ocalypse: How two critical web bugs almost broke the internet

Two unauthenticated remote code execution vulnerabilities in PHP (CVE-2026-109618 and CVE-2026-109619), dubbed PHP2Shell, were patched after researchers weaponized bugs initially dismissed in 2024. Coordinated disclosure gave major cloud providers a head start, but the vast majority of the ~70% of web servers running PHP remain exposed. The incident underscores the growing risk of AI-powered vulnerability discovery unearthing critical flaws in foundational open-source infrastructure.
General / bbtree.xiangpeng.systems

Bbtree: Fastest Ordered Map

This article introduces bbtree, a drop-in replacement for std's BTreeMap that claims to be the fastest ordered map. It explains how bbtree reduces cache misses, instructions, and cache coherence traffic, and discusses the techniques used to ensure correctness while achieving these performance gains.
[Agent Platform in Action (6)] Towards Production-Grade AI Agent Ops: Building Full-Stack Observable LINE Agents with agents-cli
AI/ML / dev.to

[Agent Platform in Action (6)] Towards Production-Grade AI Agent Ops: Building Full-Stack Observable LINE Agents with agents-cli

Recap: The Five-Stage Evolution from Local Prototype to Cloud Production-Grade AI...
Building an agentic software factory? Start with the sensor.
AI/ML / thenewstack.io

Building an agentic software factory? Start with the sensor.

A gap has long existed between the shipping velocity of the best engineering teams and the rest of the bell The post Building an agentic software factory? Start with the sensor. appeared first on The New Stack .