Skip to content

Tech News

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

Voxtral 3
AI/ML / mistral.ai

Voxtral 3

Voxtral 3 introduces state-of-the-art speech-to-text (Voxtral Small Transcribe 3) and text-to-speech (Voxtral Mini TTS 3) models alongside a voice agent SDK that enables real-time, multi-step conversations across 50+ languages. The SDK integrates listening, reasoning, tool calling, and expressive speech with emotional adaptability, while Pro Voice offers professional-grade voice cloning. Available via Mistral's EU-hosted API or self-deployment, it ensures data sovereignty for contact centers, virtual assistants, and hands-free copilots.

Why it matters

For engineers building voice-enabled AI agents, Voxtral 3 provides a production-ready pipeline combining state-of-the-art speech models with tool-calling orchestration, enabling complex, multi-step conversational workflows with data sovereignty options.
“We are not forking TypeScript”: Cloudflare spearheads Rust port written by agents, for agents
AI/ML / thenewstack.io

“We are not forking TypeScript”: Cloudflare spearheads Rust port written by agents, for agents

Cloudflare's Rust port of TypeScript, tsrs, delivers 8x speed and 70% less memory than TypeScript 7.0, with optimizations already flowing back into Microsoft's 7.1 beta (2.5x faster). Led by director of engineering Christoph Nakazawa, the effort is explicitly not a fork but a collaboration with Microsoft's TypeScript team, including principal product manager Daniel Rosenwasser, who noted that tsrs discoveries improved 7.1. Cloudflare targets AI coding agents that exhaust resources during type checking, and Rust's WebAssembly compatibility offers further advantages over Go for constrained environments.
Presentation: Beyond the Prototype: Scaling Framework Agnostic AI Agent Infrastructure with Ray
AI/ML / infoq.com

Presentation: Beyond the Prototype: Scaling Framework Agnostic AI Agent Infrastructure with Ray

The speakers discuss transitioning AI agents from local FastAPI prototypes to resilient production systems on Kubernetes using Ray. They share operational insights on RayCluster topology, stateful memory handling, auto-scaling on dynamic traffic, and robust control plane designs, while highlighting critical guardrails, multi-region setups, and LLM-as-a-judge evaluation pipelines. By Deepak Mohanakumar Chandramouli, Bhumik Vinodkumar Thakkar
Scripted tests can’t keep up with AI-written code. Datadog is betting agents can.
AI/ML / thenewstack.io

Scripted tests can’t keep up with AI-written code. Datadog is betting agents can.

Nearly a year into the agentic coding explosion, many teams now use AI to write software as their de facto The post Scripted tests can’t keep up with AI-written code. Datadog is betting agents can. appeared first on The New Stack .
Qwen3.8-Flash-Next vs Claude Opus 5.5: Six-way benchmark on a real Rust feature
AI/ML / dev.to

Qwen3.8-Flash-Next vs Claude Opus 5.5: Six-way benchmark on a real Rust feature

A controlled comparison of Qwen3.8-Flash-Next and Claude Opus 5.5 on one LlamaStash feature, across Claude Code and Pi, at medium and xhigh effort
Your Kubernetes dashboard is green. Your users are seeing red.
Cloud / thenewstack.io

Your Kubernetes dashboard is green. Your users are seeing red.

In cloud-native operations, few scenarios are as maddening as the “phantom slowdown.” This happens when Grafana or Kubernetes dashboards show The post Your Kubernetes dashboard is green. Your users are seeing red. appeared first on The New Stack .
Anthropic’s cloud partnerships are getting messy
Cloud / techcrunch.com

Anthropic’s cloud partnerships are getting messy

Anthropic's cloud partnerships with AWS, Google, and Microsoft are strained as they view the Claude maker as a competitor, forcing it to spend over $750B on cloud contracts and lease data centers from operators with limited track records. Despite an $80.7B net loss on $16.3B H1 revenue and warnings of liability for harmful AI beyond insurance coverage, Anthropic targets a November IPO with Q3 revenue up 45% to $16.8B, though two customers drive 20% of sales.
Respan spent two years building on LiteLLM. One upgrade made it start over.
AI/ML / thenewstack.io

Respan spent two years building on LiteLLM. One upgrade made it start over.

Respan released an open-source AI gateway written in Rust on Tuesday, after spending two years building more than 100,000 lines The post Respan spent two years building on LiteLLM. One upgrade made it start over. appeared first on The New Stack .
Cilium 1.20 Expands Gateway API and Opens eBPF Datapath for Kubernetes
Cloud / infoq.com

Cilium 1.20 Expands Gateway API and Opens eBPF Datapath for Kubernetes

Cilium 1.20 has been released with broader Kubernetes Gateway API support, an extensible eBPF datapath and further work on sidecarless mutual TLS. The release also makes the newer netkit network device easier to introduce across Kubernetes fleets that run different Linux kernel versions. By Matt Saunders
Paper Craft of Cracked Yellow Lock on Purple Background Front View, Data Security Concept.
General / arstechnica.com

The Internet goes on high alert after discovery of 8 compromised TLS certs

Discovery of eight eMudhra intermediate TLS certificates with NIST P-384 private keys containing only 3 to 8 bits of entropy enables anyone to mint trusted credentials for any domain. While browser vendors and major apps have already blocked the certs, cryptography engineer Filippo Valsorda warns the cleanup will surpass Heartbleed in scale, requiring updates across Linux distributions, Docker containers, unmaintained CLI tools, and IoT devices that ship their own root stores.