Episode Details

Please support this podcast by checking out our sponsors:
- KrispCall: Agentic Cloud Telephony - https://try.krispcall.com/tad
- Invest Like the Pros with StockMVP - https://www.stock-mvp.com/?via=ron
- Prezi: Create AI presentations fast - https://try.prezi.com/automated_daily


Support The Automated Daily directly:
Buy me a coffee: https://buymeacoffee.com/theautomateddaily

Today's topics:

US export controls hit Anthropic - A US export-control directive forces Anthropic to suspend Fable 5 and Mythos 5 access for all customers, raising major national-security, transparency, and deployment-precedent questions.

OpenAI vs Anthropic memory - A new comparison argues OpenAI leans on server-side context compaction while Anthropic prefers multi-agent delegation—two competing approaches to long, complex tasks that may converge.

GitHub goes multi-cloud under load - Microsoft is reportedly adding AWS capacity to stabilize GitHub after agentic coding activity drove outages, signaling real-world limits to single-cloud migration plans.

AI code quality vs incidents - New Relic’s 2026 report finds AI-generated code can look fine in review but correlates with more production incidents, pushing observability, testing, and governance to the forefront.

Siri may route to rivals - iOS 27 beta leaks suggest Siri could switch between third-party models like ChatGPT, Claude, and Gemini—turning Siri into a routing layer amid EU DMA pressure and partnership tension.

Open Knowledge Format for agents - Google Cloud’s Open Knowledge Format (OKF) proposes a vendor-neutral ‘LLM wiki’ bundle in Markdown and YAML, aiming to make org context portable for AI agents and teams.

Sparse attention speeds long context - MiniMax open-sourced Sparse Attention kernels targeting next-gen NVIDIA GPUs, a sign that long-context performance is increasingly won with smarter attention, not just bigger chips.

Inference cost hinges on memory - A ‘napkin math’ cost model says long-context inference is often memory-bandwidth and KV-cache constrained, explaining why batching, paging, and cache management drive profitability.

Europe’s federated sovereign compute - The euromesh project claims Europe could train a sovereign frontier model sooner by federating existing EuroHPC and public compute, though politics and scheduling may be the hard part.

Tooling for safer agent workflows - Strands Agents released an open-source agent harness with hooks, tracing, and guardrails—reflecting the push for controllable, observable agent behavior across model providers.

Continuous LLM evaluation in practice - AllenAI’s olmo-eval targets the everyday loop of comparing checkpoints with reproducible suites and statistical signal checks, making evaluation less ‘leaderboard’ and more ‘engineering.’

-OpenAI vs Anthropic: Compaction vs Sub-Agent Delegation for Long-Context Work
-MiniMax open-sources MSA sparse attention and FlashAttention kernels for NVIDIA SM100
-Report Claims Europe Could Train a Frontier AI Mod