The New Stack
@thenewstack.io
All about at-scale software development, deployment & management. Tech news, analysis, research, podcasts, videos & more! Website 🌐 https://thenewstack.io Write for TNS ✍️ https://thenewstack.io/contributions Subscribe 📩 https://thenewstack.io/newsletter
AI-generated bug reports are flooding Apple's security pipeline, prompting new submission caps that could slow down legitimate vulnerability research.
Apple's bug bounty limit reveals a problem no one knows how to solve yet
AI-generated bug reports are flooding Apple's security pipeline, prompting new submission caps that could slow down legitimate vulnerability research.
bit.ly
Nscale's Anyscale acquisition raises lock-in questions as the GPU neocloud builds what its CEO calls the first full-stack AI hyperscaler.
Nscale just bought Anyscale. Here's why it matters for multi-cloud neutrality.
Nscale's Anyscale acquisition raises lock-in questions as the GPU neocloud builds what its CEO calls the first full-stack AI hyperscaler.
bit.ly
Alibaba's Qwen3.8-Max spent 16 days building a CLI tool autonomously — 265 commits, 127 PRs, zero human help. The GitHub audit trail is public.
Alibaba's AI coded for 16 days straight and every commit is on GitHub
Alibaba's Qwen3.8-Max spent 16 days building a CLI tool autonomously — 265 commits, 127 PRs, zero human help. The GitHub audit trail is public.
bit.ly
AI code generation shifts focus to platform engineering. Discover why software companies must build internal tools to harness AI safely.
Every software company will become a dev tools company
AI code generation shifts focus to platform engineering. Discover why software companies must build internal tools to harness AI safely.
bit.ly
Google DeepMind loses Jeff Dean and three top engineers to Discovery Loop, a startup automating scientific research, as Demis Hassabis moves to a strategy role.
Google’s four AI departures: "We wanted to build something differently"
Google DeepMind loses Jeff Dean and three top engineers to Discovery Loop, a startup automating scientific research, as Demis Hassabis moves to a strategy role.
bit.ly
Kief Morris explains how CI/CD and harness engineering keep AI-generated software safe and secure, production-ready and aligned with real-world user needs.
Forget humans “in” the loop. Harness engineering puts humans “on” the loop.
Kief Morris explains how CI/CD and harness engineering keep AI-generated software safe and secure, production-ready and aligned with real-world user needs.
bit.ly
Meta is turning its engineers' code fixes into training data for its AI coding agent, MetaCode, as it races to close the gap with OpenAI and Anthropic.
The 800 mistakes that could reshape Meta's AI coding strategy
Meta is turning its engineers' code fixes into training data for its AI coding agent, MetaCode, as it races to close the gap with OpenAI and Anthropic.
bit.ly
Google DeepMind’s three new models give robots full-body control, greater dexterity, and reasoning capabilities to execute multi-step tasks
Gemini Robotics 2 brings us one step closer to physical AGI
Google DeepMind’s three new models give robots full-body control, greater dexterity, and reasoning capabilities to execute multi-step tasks
bit.ly
Anthropic found three cases where Claude models escaped test sandboxes and compromised real systems, exposing why AI evaluation infrastructure needs production-grade security.
What Claude’s real-world breaches reveal about AI safety tests
Anthropic found three cases where Claude models escaped test sandboxes and compromised real systems, exposing why AI evaluation infrastructure needs production-grade security.
bit.ly
DeepSeek has launched V4-Flash-0731, using post-training to boost agent performance while keeping the same architecture and offering open weights.
DeepSeek's smaller model just outperformed its own flagship
DeepSeek has launched V4-Flash-0731, using post-training to boost agent performance while keeping the same architecture and offering open weights.
bit.ly
Fresh off a $300 million round, Abbas turned Temporal into a test case for agentic AI, shipping features 20% to 30% faster. Whether that speed reaches customers as value, or the business is just growing alongside it, is what he can't yet answer.
Temporal 5x’d AI spend and doubled revenue. CEO Samar Abbas says he can’t prove they’re connected.
Fresh off a $300 million round, Abbas turned Temporal into a test case for agentic AI, shipping features 20% to 30% faster. Whether that speed reaches customers as value, or the business is just growing alongside it, is what he can't yet answer.
bit.ly
OpenAI says an internal model produced 10 new math proofs for about $2,000 in API tokens, reframing frontier AI reasoning as a budgetable expense.
OpenAI's Astra just proved 10 long-standing math and science theorems. The tokens cost $2,000.
OpenAI says an internal model produced 10 new math proofs for about $2,000 in API tokens, reframing frontier AI reasoning as a budgetable expense.
bit.ly
OpenAI has lowered API prices for two GPT-5.6 models only three weeks after their launch. On Thursday, the company announced
OpenAI slashes API costs amid rising global competition
OpenAI has lowered API prices for two GPT-5.6 models only three weeks after their launch. On Thursday, the company announced
bit.ly
Thibault Sottiaux, who leads core products at OpenAI, believes that today’s version of Codex will seem outdated before the year
Today's Codex will feel "primitive" by fall — and its own team's roadmap backs it up
Thibault Sottiaux, who leads core products at OpenAI, believes that today’s version of Codex will seem outdated before the year
bit.ly
Alibaba's Qwen3.8-Max claims frontier scale and open weights, but developers say the benchmarks, context window and 16-day agent run deserve scrutiny.
Alibaba Qwen3.8-Max reactions: "An API business model wearing an open source jacket"
Alibaba's Qwen3.8-Max claims frontier scale and open weights, but developers say the benchmarks, context window and 16-day agent run deserve scrutiny.
bit.ly
Experts say NOOA, the open-source Python framework, could make agents easier to test and inspect but raises new questions about review and security.
Nvidia's NOOA makes an agent one Python class
Experts say NOOA, the open-source Python framework, could make agents easier to test and inspect but raises new questions about review and security.
bit.ly
OpenAI and Elastic have deepened their partnership to give enterprise AI agents secure context, better observability, and quicker, evidence-based threat response.
OpenAI and Elastic are tackling the AI problem enterprises can’t ignore
OpenAI and Elastic have deepened their partnership to give enterprise AI agents secure context, better observability, and quicker, evidence-based threat response.
bit.ly
6 months after Cloudflare acquired the team behind Astro, co-founder Fred Schott explains how an AI agent is emptying the project's issue tracker -- and why the hardest part is still left to a human
Astro's GitHub issue backlog is heading to zero for the first time in 5 years. Now Cloudflare is open-sourcing the tool that did it.
6 months after Cloudflare acquired the team behind Astro, co-founder Fred Schott explains how an AI agent is emptying the project's issue tracker -- and why the hardest part is still left to a human
bit.ly
AI-written code is outpacing security controls, pushing platform teams toward AI bills of materials, runtime shields and faster, patchable platforms today.
AI-generated software is forcing yet another platform rethink
AI-written code is outpacing security controls, pushing platform teams toward AI bills of materials, runtime shields and faster, patchable platforms today.
bit.ly
OpenAI says an internal model produced 10 new math proofs for about $2,000 in API tokens, reframing frontier AI reasoning as a budgetable expense.
OpenAI's Astra just proved 10 long-standing math and science theorems. The tokens cost $2,000.
OpenAI says an internal model produced 10 new math proofs for about $2,000 in API tokens, reframing frontier AI reasoning as a budgetable expense.
bit.ly
YugabyteDB AMP uses serverless PostgreSQL and four AI agents to manage database fleets, scale idle workloads to zero and support scaling without migration.
AI agents can create database sprawl issues. YugabyteDB’s solution is more agents!
YugabyteDB AMP uses serverless PostgreSQL and four AI agents to manage database fleets, scale idle workloads to zero and support scaling without migration.
bit.ly
Nscale's Anyscale acquisition raises lock-in questions as the GPU neocloud builds what its CEO calls the first full-stack AI hyperscaler.
Nscale just bought Anyscale. Here's why it matters for multi-cloud neutrality.
Nscale's Anyscale acquisition raises lock-in questions as the GPU neocloud builds what its CEO calls the first full-stack AI hyperscaler.
bit.ly
Thibault Sottiaux, who leads core products at OpenAI, believes that today’s version of Codex will seem outdated before the year
Today's Codex will feel "primitive" by fall — and its own team's roadmap backs it up
Thibault Sottiaux, who leads core products at OpenAI, believes that today’s version of Codex will seem outdated before the year
bit.ly
Kief Morris explains how CI/CD and harness engineering keep AI-generated software safe and secure, production-ready and aligned with real-world user needs.
Forget humans “in” the loop. Harness engineering puts humans “on” the loop.
Kief Morris explains how CI/CD and harness engineering keep AI-generated software safe and secure, production-ready and aligned with real-world user needs.
bit.ly
Google DeepMind’s three new models give robots full-body control, greater dexterity, and reasoning capabilities to execute multi-step tasks
Gemini Robotics 2 brings us one step closer to physical AGI
Google DeepMind’s three new models give robots full-body control, greater dexterity, and reasoning capabilities to execute multi-step tasks
bit.ly
AI researchers and top lab executives sign open letter urging the U.S. to build a legal framework for slowing AI development before safety falls behind
Anthropic backs urgent call for the most powerful AI labs to hit the brakes
AI researchers and top lab executives sign open letter urging the U.S. to build a legal framework for slowing AI development before safety falls behind
bit.ly
Alibaba's Qwen3.8-Max claims frontier scale and open weights, but developers say the benchmarks, context window and 16-day agent run deserve scrutiny.
Alibaba Qwen3.8-Max reactions: "An API business model wearing an open source jacket"
Alibaba's Qwen3.8-Max claims frontier scale and open weights, but developers say the benchmarks, context window and 16-day agent run deserve scrutiny.
bit.ly
PortSwigger launches Burp AT, agentic AI pentesting where agents propose actions but a deterministic control layer enforces scope and approval.
"The beast needs a cage": Why PortSwigger's agentic pentesting is kept safe behind bars
PortSwigger launches Burp AT, agentic AI pentesting where agents propose actions but a deterministic control layer enforces scope and approval.
bit.ly
Modus has emerged from stealth to build a context warehouse that gives AI agents only the enterprise information they need.
Modus raises $10M to give AI agents only what they need — and nothing more.
Modus has emerged from stealth to build a context warehouse that gives AI agents only the enterprise information they need.
bit.ly
Experts say NOOA, the open-source Python framework, could make agents easier to test and inspect but raises new questions about review and security.
Nvidia's NOOA makes an agent one Python class
Experts say NOOA, the open-source Python framework, could make agents easier to test and inspect but raises new questions about review and security.
bit.ly