AI Coding Live News

OpenAI Daybreak Red (GPT-5.6 Cyber) arrives on Amazon Bedrock in AWS GovCloud (US-East) for vetted defenders, at $16.875 input and $101.25 output per 1M tokens

On Oct 30, AWS announced that Daybreak Red (GPT-5.6 Cyber, model ID openai.gpt-5.6-cyber), the OpenAI security model built for vulnerability research, is now on Amazon Bedrock in AWS GovCloud (US-East), after previously being offered only in US East (Ohio). Access still requires passing OpenAI Daybreak eligibility. GovCloud short-context pricing is $16.875 input and $101.25 output per 1M tokens, about 23% above Ohio at $13.75 and $82.50.

Key Takeaways

  • Regions: Daybreak Red can now be called In-Region in us-east-2 (Ohio) and us-gov-east-1 (GovCloud US-East); Geo and Global cross-Region inference are not supported (AWS announcement)
  • Eligibility: customers must first enroll in OpenAI Daybreak / Trusted Access for Cyber, then have their AWS account team enable access
  • Pricing: GovCloud short context (272K input tokens or fewer) is $16.875 input, $21.09375 for 30-minute cache writes, $1.6875 cache reads and $101.25 output per 1M tokens
  • Data: AWS says inference data is not used for training and Daybreak Red data does not need to be shared with OpenAI; it runs on an inference engine with zero-operator access enforced at the chip
  • API: served through the bedrock-mantle endpoint with OpenAI Responses and Chat Completions, and a 1,050,000-token context window
Read full story

Alibaba Cloud opens a Netherlands region at its first international Apsara Conference, plans agentic AI services, and JetBrains explores Qwen for Junie

On Oct 29, Alibaba Cloud held its first international Apsara Conference in Amsterdam and launched its first public cloud region in the Netherlands, with two data centres, making it the company's fourth region in Europe. It plans to bring agentic AI services to the region, including ACS Agent Sandbox, Agent Security Center, AI Security Guardrails 2.0, Agentic SOC and STAROps, and unveiled a new generation of its Apsara Stack hybrid cloud. On the partner side, JetBrains is working with Alibaba Cloud to explore Qwen models for its Junie coding assistant, in both cloud and local setups.

Alibaba Cloud opens a Netherlands region at its first international Apsara Conference, plans agentic AI services, and JetBrains explores Qwen for Junie

Key Takeaways

  • The Netherlands region has 2 data centres and is Alibaba Cloud's 4th in Europe after Frankfurt, London and Paris; the global footprint is now 32 regions and 109 availability zones (company figures)
  • Five agentic AI services are planned for the region: ACS Agent Sandbox (hardware-level isolation), Agent Security Center, AI Security Guardrails 2.0, Agentic SOC and STAROps
  • The new Apsara Stack offers 80+ cloud and AI products; Alibaba says AI-driven operations reach 92.5% accuracy and a 70% automation rate
  • JetBrains is exploring Qwen models for Junie, running either in-region in the cloud or fully local; orq.ai, Monks, Magnific and Idea.ai also integrate Alibaba models
  • Over the next 12 months Alibaba plans its first region in Finland and more data centre capacity in Germany and France, as part of its previously announced US$53B AI and cloud investment
Read full story
Linked Models & Agents:

TechCrunch/The Information: Astra uses recurrent-depth reasoning; safety experts warn CoT monitoring gets harder

Per The Information, reported by TechCrunch and The Verge, OpenAI's upcoming Astra uses “recurrent depth” (also called opaque recurrence / a looped transformer): cycling computation through internal layers before output, leaving fewer legible natural-language chain-of-thought traces. That can boost performance while making misbehavior harder to monitor. Redwood's Ryan Greenblatt called it potentially “the single worst development for AI security/safety to date”; Zvi Mowshowitz and others warned of a race to the bottom on monitorability. OpenAI says use is limited and CoT remains legible; chief scientist Jakub Pachocki stressed CoT monitoring and said Astra's compute depth is within about a factor of two of GPT-4. TechCrunch says a Wednesday-morning Information follow-up reported Anthropic and Google DeepMind are already discussing the technique.

Key Takeaways

  • Astra reportedly uses recurrent depth / opaque recurrence, harder-to-read internal reasoning
  • Safety researchers fear a monitorability race to the bottom; OpenAI says use is limited with CoT monitoring
  • Follow-up: Anthropic and DeepMind are already discussing the technique
Read full story