"NUNTII EX MACHINA"
DOCUMENTING THE RACE TO AGI

THE
AGENTIC
TIMES_

▮ NEWS_TICKER.LIVE REC
▸ AI model escapes and hacks another company during testing thenationalnews.com · 22.07.2026▸ Tech Briefs: Virginia Tech researchers’ AI model is meant to manage water use Cardinal News · 22.07.2026▸ XRP Ledger Surpasses 1 Million AI Agent Transactions Coinpedia · 22.07.2026▸ OpenAI's secret model escapes test, breaches another AI model Türkiye Today · 22.07.2026▸ Three AI Agents Held a Conference and Decided I Should Do the Work Instead HackerNoon · 22.07.2026▸ US' allegations against Chinese AI models reflect anxiety over China's AI rise, risk undermining… Global Times · 22.07.2026▸ How I Built an AI Agent That Makes Videos While I Sleep HackerNoon · 22.07.2026▸ GitLab 19.2 lets AI agents resolve vulnerabilities on their own Techzine Global · 22.07.2026▸ OpenAI says AI models autonomously pulled off a major hack, but only a Chinese… Digital Trends · 22.07.2026▸ OpenAI says AI model hacked another company's systems during internal test Fox Business · 22.07.2026▸ OpenAI’s AI agent broke out of a test environment and hacked Hugging Face calcalistech.com · 22.07.2026▸ Xometry updates AI models for custom manufacturing Engineering.com · 22.07.2026▸ How Security AI Agents can boost Corporate Networks Cybersecurity Insiders · 22.07.2026▸ OpenAI’s AI agent goes rogue, hacks Hugging Face’s internal systems The Hindu · 22.07.2026▸ The Agentic AI Mirage: Why Your ‘Personalized’ Assistant is Working for the Vendor, Not… CustomerThink · 22.07.2026▸ The new ad for chips? An AI model Breaking Bad TradingView · 22.07.2026▸ OpenAI reveals AI models go rogue during testing: Here’s what happened The News International · 22.07.2026▸ Days after world’s biggest AI models repository Hugging Face announced that it has been… The Times of India · 22.07.2026
▮ MODEL_FEED.LIVE REC
▸ Kimi-K2.7-Code-DFlash NVIDIA · coding · 08.07.2026▸ Nemotron-Labs-Audex-30B-A3B NVIDIA · 30B · MoE · 06.07.2026▸ Nemotron-Labs-Audex-2B NVIDIA · 2B · 06.07.2026▸ DeepSeek-V4-Flash-DSpark DeepSeek · Open language model · 27.06.2026▸ Qwen-AgentWorld-35B-A3B Qwen · 35B · MoE · multimodal · 22.06.2026▸ GLM-5.2 Zhipu · Open language model · 16.06.2026▸ North-Mini-Code-1.0 Cohere · coding · 05.06.2026▸ DeepSeek-V4-Pro DeepSeek · Open language model · 22.04.2026▸ DeepSeek-V4-Flash DeepSeek · Open language model · 22.04.2026▸ granite-4.1-8b IBM · 8B · 06.04.2026▸ granite-4.1-3b IBM · 3B · 06.04.2026▸ Mamba2-primed-HQwen3-8B-Instruct Amazon · 8B · instruct · 31.03.2026▸ Falcon-OCR TII · Open language model · 22.02.2026▸ tiny-aya-base Cohere · Open language model · 13.02.2026▸ tiny-aya-global Cohere · Open language model · 13.02.2026▸ GLM-5 Zhipu · Open language model · 11.02.2026
▮ INDUSTRY / 29.04.2026 · 2 MIN READ

AWS and OpenAI Go All in With Managed Agents

AWS and OpenAI Go All in With Managed Agents

This week the two companies expanded their multi-year, $38 billion partnership with the announcement that Amazon Bedrock Managed Agents, powered by OpenAI, is now available in limited preview for agent builders. They also announced that GPT-5.4, GPT-5.5, and OpenAI’s Codex coding agent are also accessible via Amazon Bedrock.

The service is built around what OpenAI calls the ‘agent harness’, the orchestration layer that transforms a foundation model into a functioning agent capable of reasoning, planning, and completing multi-step tasks. In practice, this has been one of the hardest parts of enterprise agent deployment to get right. That’s because building and maintaining the harness requires substantial engineering effort, and organisations frequently get it wrong in ways that create performance, security, or reliability problems, sometimes all of them at once!

Bedrock Managed Agents handles this infrastructure entirely within AWS. Each agent runs with its own identity, logs every action for auditability, and executes inside the customer’s own environment, with model inference staying on Amazon Bedrock and data never leaving the cloud. The service connects directly into AWS compute and data services, and is designed to work alongside Bedrock AgentCore, AWS’s broader open platform for building and orchestrating agent fleets.

The practical effect, as analysts have noted when the news broke, is that enterprises no longer need to resolve the underlying model selection question before building. The architectural decisions have been made. Teams can start with the agent and trust that the infrastructure beneath it meets enterprise security and governance requirements from day one, a big barrier has been cleared for many already looking at the cloud.

Microsoft’s Azure, long assumed to be the natural home for OpenAI’s commercial ambitions, does not look so monopolistic now. AWS, with its substantially larger enterprise footprint and deeper penetration into production infrastructure, becomes an equal, if not primary, deployment channel. The agent war just opened up a new front, but it’s far from over.

END OF TRANSMISSION ▮ ◂ MORE INDUSTRY