In Season 5, Episode 16, Karl and Jon are joined by Taylor Dolezal, Head of Open Source at Dosu, to discuss AWS Lambda’s S3 file system integration, Anthropic Claude Opus 4.7 arriving in Amazon Bedrock, Amazon Q’s cost management capabilities for FinOps, Microsoft’s £2.8 billion UK licensing lawsuit, and the latest AI adoption statistics and implementation challenges, before taking a nostalgic tangent into early-noughties personal digital assistants...
03:13 - AWS Lambda Mounts S3 Buckets as File Systems
AWS Lambda can now mount S3 buckets as file systems via S3 Files, reducing the need for separate EFS infrastructure when working with large files already in S3. This simplifies use cases like image processing, user-generated content, Lambda Durable Functions, and multi-step AI workflows.
Jon noted that while the feature reduces engineering complexity, pricing remains hard to predict. It also does not resolve some S3 limitations, such as folder recreation issues, and its usefulness depends on Lambda capacity provider configuration.
11:47 - Claude Opus 4.7 Available in Amazon Bedrock
Anthropic’s Claude Opus 4.7 is now available in Amazon Bedrock, continuing Anthropic’s rapid model release cycle. The model offers new capabilities, but users may need to adjust prompts and implementations to see improvements.
The discussion noted that newer models do not always outperform older ones in real-world use. Some teams still prefer Claude 4.6 for reliability. Taylor highlighted that performance can vary by infrastructure, such as Nvidia chips versus TPUs, and that enterprise tools for evaluating model cost, consistency, and behavior remain immature. Concerns were also raised about models detecting test environments and optimizing for metrics in unintended ways.
20:19 - Amazon Q Adds FinOps Cost Management
Amazon Q is now integrated into AWS Cost Explorer, letting users ask natural language questions about AWS spending without manually building reports. This makes cost insights more accessible to business users and non-technical stakeholders.
Taylor emphasized that cost predictability is critical for planning, based on experience at the Linux Foundation and Disney. While the feature lowers the barrier to cost analysis, it currently focuses on ad-hoc questions rather than recurring reports, dashboards, or automated alerts. Jon suggested future improvements such as monthly recurring analysis and natural language dashboard creation.
26:21 - Microsoft Faces UK Cloud Licensing Lawsuit
Microsoft is facing a £2.8 billion UK lawsuit alleging anti-competitive cloud licensing practices, claiming Microsoft software such as Windows Server and SQL Server costs more on rival clouds than on Azure.
Jon noted that while Azure pricing advantages may be expected, Microsoft has not clearly explained how price differences are calculated. Microsoft argues damages are difficult to assess because cloud providers do not separate licensing from infrastructure costs. Taylor shared that Disney had to migrate SQL Server workloads after Microsoft restricted bring-your-own-license options, creating major engineering effort. The case could force more transparency in cloud licensing, though changes would likely take time.
34:12 - AI Adoption Still Mostly Basic
AWS research presented at the London Summit found that 64% of UK organizations have adopted AI, but only 25% of those use it at an advanced level, equal to about 15% of all businesses.
The discussion questioned what counts as “basic” versus “advanced” AI use, such as whether Office 365 Copilot or code generation qualifies. Jon stressed the importance of definitions. Taylor argued that the biggest barriers are governance, compliance, and unclear use cases rather than skills alone. Key challenges include data residency, government requirements, supply chain security, rapid technology change, and likely future cost increases as VC subsidies decline.
44 min