OpenAI's enterprise AI stack gets specific
OpenAI is packaging Frontier, ChatGPT Enterprise, Codex, and AWS runtime as one enterprise stack, with CyberAgent as the clearest proof point yet.
Signed archive
Idris Vale
Idris writes about the institutional machinery around AI, but the lens is broader than policy alone: procurement frameworks, public-sector buying rules, platform leverage, compliance burdens, workflow risk, and the market structure hiding beneath product or infrastructure headlines. The through-line is practical power, not abstract theater.
Brussels · London corridorFollow the buying process, not just the bill text.Background in public-sector market analysis and regulatory briefings.
Back to all authors
Latest story
Claude Cowork gets the enterprise control layer
Anthropic's Apr. 9 update gives Claude Cowork the RBAC, spend caps, analytics, and telemetry that turn a flashy desktop agent into something companies can actually roll out.
- 23
- Apr 10, 2026
- 5
Coverage signature
Policy turns real when someone has to buy the system.
Institutional, commercial, and allergic to vague compliance talk.
Coverage lanes
Institutional adoptionProcurementRegulationMarket structure
EU AI procurementAI market structurepublic sector AI buying
Published stories
Everything currently attached to this byline.
Safetensors joins PyTorch Foundation weight stack
Hugging Face has moved Safetensors into the PyTorch Foundation, turning the default open-model weight format into vendor-neutral infrastructure for the wider stack.
GLM-5.1 coding agent and the 8-hour shift
Z.AI pitches GLM-5.1 as an 8-hour coding agent and documents manual routes into Claude Code, OpenClaw, and OpenRouter. Access is real. Proof is pending.
Physical AI is leaving the lab, and NVIDIA wants in
NVIDIA's National Robotics Week roundup linked household research, startup pipeline, and solar-field deployment into one bid to own the platform layer under physical AI.
Anthropic locks in Google TPU capacity with Broadcom
Anthropic's Google Cloud and Broadcom pact is a compute-capacity story, not a model launch. It shows frontier AI shifting toward power, silicon, and reserved compute.
GLM-5.1 hits Hugging Face. Now the scrutiny starts
Z.ai's MIT-licensed GLM-5.1 and GLM-5.1-FP8 checkpoints hit Hugging Face with vLLM and SGLang support, so the benchmark story now has to survive deployment reality.
Gemini 3 is Google's agent stack, not one model
Gemini 3 matters because Google tied 3.1 Pro, Flash, Flash-Lite, Flash Live, AI Studio, and Antigravity into one app-building stack for developers.
Reuters says DeepSeek V4 will run on Huawei chips
Reuters says DeepSeek V4 will run on Huawei chips. If true, the bigger story is China moving a flagship AI cycle onto a homegrown silicon and software stack.
GitHub Copilot SDK exports GitHub's agent runtime
GitHub's Copilot SDK public preview is not just another toolkit. It exports the agent runtime behind Copilot CLI and cloud agent into apps.
OpenAI audio models plug the missing agent voice layer
OpenAI next-generation audio models look less like a model drop and more like the missing voice layer for its agent stack, with better transcription, steerable TTS, and live voice-agent docs.
Google Vids turns Veo into a free entry point
Google Vids just got Veo 3.1 for any Google account, but the bigger story is the bundle: free clips, Chrome recording, YouTube export, and a tier chart with trust issues.
AWS puts frontier agents on security and ops duty
AWS has tied its frontier-agents pitch to two GA products: one for autonomous penetration testing, one for incident response across AWS and beyond.
