Please confirm you are human

This browser or connection looks automated. Press and continuously hold the control for 3 seconds to enable Google-hosted web results and, when separately allowed, AI-assisted answers.

A successful check enables 100 search requests. Interactive access does not authorize scraping, systematic collection, or reuse of search output.

Hold with a pointer, or hold Space or Enter.

News

The New Stack
thenewstack.io > openai-gpt56-cyber-daybreak

OpenAI built a model it doesn't want most people to use

1+ hour, 22+ min ago   (369+ words) OpenAI's new GPT-5.6 Cyber model handles exploit chains and zero-day hunting that its general-purpose models block, available through the gated Daybreak Red tier....

The New Stack
thenewstack.io > meta-glimmer-distillation-agents

Meta's Muse Glimmer fits on a laptop

5+ hour, 21+ min ago   (813+ words) Meta released Muse Glimmer on Monday, a 30-billion-parameter open-weight model designed to run agentic workflows on local hardware. It’s available for download on Hugging Face, but even more notable is how Meta turned its larger Muse Spark model into something…...

The New Stack
thenewstack.io > accelerating-eks-image-pulls

Pulling multi-gigabyte container images in seconds on Amazon EKS

6+ hour, 37+ min ago   (514+ words) Speed up large ML container image pulls on Amazon EKS from minutes to seconds using parallel download and unpack in containerd....

The New Stack
thenewstack.io > meta-muse-claude-code

Meta Muse Code vs. Fable 5: Meta Muse is cheaper, but at what cost?

7+ hour, 37+ min ago   (668+ words) Meta admits its first coding agent doesn't perform as well as Claude Code. I ran both on three real coding jobs and tracked every token. Muse’s refactor barely restructured anything, and it left dead code behind....

The New Stack
thenewstack.io > developers-review-gpt-56-sol

"It blows my mind”-“It has a tendency to overengineer things a little”: Developers react to road-testing OpenAI GPT‑5.6 Sol

8+ hour, 9+ min ago   (381+ words) Developers weigh in on OpenAI's GPT-5.6 Sol, from database audits to Erdős problem solving, as it takes on Anthropic's Claude Opus 5....

The New Stack
thenewstack.io > deepseek-flash-pro-benchmark

V4-Flash vs. V4-Pro: DeepSeek promised better and cheaper. It's true, but not how I expected.

10+ hour, 37+ min ago   (513+ words) DeepSeek says its refreshed V4-Flash beats its V4-Pro preview on coding at a third of the price. I ran both on three real coding jobs and tracked every token. Flash really was better but used more tokens, making the bills…...

The New Stack
thenewstack.io > real-cost-diy-platform

Platform Engineering ROI: What it costs to build your own platform

1+ day, 6+ hour ago   (209+ words) Building a custom developer platform costs $7.5M a year. Discover the hidden ROI of buying versus building internal platforms....

The New Stack
thenewstack.io > evaluating-coding-agents-framework

Coding agents can be evaluated. We just have to evaluate the work.

1+ day, 7+ hour ago   (546+ words) Stop grading coding agents like chatbots. Here is how to evaluate non-deterministic AI agents using executable contracts and scorecards....

The New Stack
thenewstack.io > ai-productivity-measurement-gap

AI coding got faster. Why didn’t engineering?

1+ day, 8+ hour ago   (907+ words) AI is great at making individuals faster, but the surrounding systems are then slowing everything right back down. This result — or, rather, lack thereof — is amplified by company size and pull request size. To the point that, while AI investment…...

The New Stack
thenewstack.io > ai-adoption-versus-usage

AI adoption isn’t the same as AI usage

2+ day, 7+ hour ago   (579+ words) High token spend isn't real AI adoption. Discover how engineering teams can move past vanity metrics to create lasting workflow shifts....