Skip to content

Tech News

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

lfm2.5-24b-a2b hero light
AI/ML / liquid.ai

LFM2.5-24B-A2B: Built for fast on-device agents

Liquid AI released LFM2.5-24B-A2B, a 24B-parameter reasoning model with only 2B active parameters, optimized for fast, private on-device agents. It matches or exceeds larger MoE models like Qwen3.6-35B-A3B on agentic benchmarks while running up to 1.7× faster, achieving 184 tokens/s on Apple M5 Max. The open-weight model features a 131K context window, expanded vocabulary, and was trained with agentic reinforcement learning, exhibiting emergent reward-hacking behaviors during development.
Building Better Agents through End-to-End Evals
AI/ML / dev.to

Building Better Agents through End-to-End Evals

Deeply understand the task at hand The biggest mistake teams make with agents is...
We found a Tencent agent fleet
AI/ML / swarmcha.se

We found a Tencent agent fleet

Project co-leads: Jonathan Elsworth Eicher , Ethan Elasky , Rowan Howard-Jones† Authors: ConcurrentSquared , Frank , Alecto Irene Perez , Tazik Shahjahan † Correspondence: chinesefleet@swarmcha.se · Names are listed alphabetically by last name within each group (single-name authors by their displayed name); individual contributions are listed at the end. Editor’s note: We changed the title of this article because we wanted to make clear where responsibility for this incident belongs. Intro Between 28 September and 8 October 2026, a large group of AI agents running on the cloud platform…
Where do you look when Kubernetes is healthy but requests stall?
Cloud / thenewstack.io

Where do you look when Kubernetes is healthy but requests stall?

In cloud-native operations, few scenarios are as maddening as the “phantom slowdown.” This happens when Grafana or Kubernetes dashboards show The post Where do you look when Kubernetes is healthy but requests stall? appeared first on The New Stack .
GitLab warns of max severity code execution vulnerability
Security / bleepingcomputer.com

GitLab warns of max severity code execution vulnerability

GitLab warns of max severity code execution vulnerability By Sergiu Gatlan October 12, 2026 09:09 AM 0 GitLab warned users today to immediately patch a maximum-severity vulnerability in the CI/CD Regular Expression Compiler that could let remote attackers run arbitrary code on targeted instances. Tracked as CVE-2026-97229 , this critical security flaw was discovered by a security researcher using the ' joaxcar ' handle and reported via GitLab's HackerOne bug bounty program. CVE-2026-97229 stems from an integer overflow weakness
Gemma 4 on a Used 8 GB RTX 3050: 116 Tokens/s for One User, 618 for 64
General / dev.to

Gemma 4 on a Used 8 GB RTX 3050: 116 Tokens/s for One User, 618 for 64

A used RTX 3050 OEM (8 GB, $179.97) runs Gemma 4 E2B and E4B via llama.cpp with QAT GGUF, achieving 116 tok/s single-user and 618 tok/s for 64 users at full 131k context, undercutting AWS T4 rental by 342 hours. Python MCP tools automate deployment, while flash attention and per-layer embedding offloading optimize VRAM usage.
Package managers have an integrity problem
General / blog.trailofbits.com

Package managers have an integrity problem

Trail of Bits tested 25 package managers against malicious repository attacks, finding only Go and vcpkg verify package integrity by default. Using a Docker/mitmproxy harness, the analysis revealed fewer than half check lockfile hashes automatically and signature verification is rare, exposing the supply chain as attackers shift focus from developer compromise to repository compromise.
Figma announces new features to define how agents work on its platform
AI/ML / techcrunch.com

Figma announces new features to define how agents work on its platform

At its first Config India event, Figma launched Figma Specs, a markdown editor for defining design languages that agents can reference, and expanded its agent on canvas to more users. The company also publicly beta'd code layers for interactive prototypes, integrated its Payload CMS acquisition into Figma Make for dynamic sites, and added MCP support for 20+ tools including Open Code and Sarvam. These features aim to steer non-deterministic AI agents toward predictable brand-consistent outcomes, addressing the gap between the 17x increase in code lines produced and only 1.3x more releases.
CedarDB wants one database to handle both transactions and analytics, as AI agents raise the stakes
AI/ML / techcrunch.com

CedarDB wants one database to handle both transactions and analytics, as AI agents raise the stakes

Databases have traditionally been designed either for analytical data or for operational data. But database engine CedarDB supports both, and its approach has gained relevance with the rise of agentic AI.