The mythos of Mythos and Allbirds takes flight to the neocloud

What It Covers

Practical AI hosts Daniel Whitenack and Chris Benson cover four converging developments: Allbirds' pivot from footwear to AI compute infrastructure, the emergence of NeoCloud providers like CoreWeave, Anthropic's unreleased Mythos model and its security implications, and the legal risks of using AI chatbots for confidential business communications.

Key Questions Answered

• •NeoCloud Infrastructure: AI-native cloud providers like CoreWeave, Together AI, and Lambda Labs build GPU-first infrastructure specifically for training and inference workloads, not general computing. This specialization exists because GPU supply is constrained within hyperscalers like AWS, and AI workloads require massive multi-node data movement patterns that general-purpose cloud architecture does not efficiently support.
• •Allbirds Capital Reallocation: Allbirds sold its entire footwear operation to American Exchange Group in March 2026, retaining roughly $50M in cash plus a public stock ticker, then pivoted to AI compute infrastructure. Shares jumped approximately 700% on the announcement. The critical limitation: $50M is a marginal entry point in a market where competitors invest billions.
• •Token Maxing as Vanity Metric: Companies like Meta have created internal leaderboards tracking developer token consumption to measure AI-assisted productivity, but token spend functions like website clicks — a correlation with output, not a cause. Developers game leaderboards by spending tokens on trivial tasks. The actual productivity-to-cost ratio across organizations remains unmeasured and highly variable.
• •Mythos Security Implications: Anthropic's unreleased Mythos model reportedly identified thousands of vulnerabilities across every major operating system and browser. Rather than releasing it publicly, Anthropic launched Project Glasswing, a closed program with approximately 40 companies given access to audit and patch their systems. This signals a broader tailwind for AI governance, security operations, and AI underwriting companies.
• •AI Chat and Attorney-Client Privilege: A federal judge ruled that Claude chat logs used to prepare legal materials are discoverable in court because AI systems are classified as third parties, not attorneys. Pasting confidential contracts, draft agreements, or sensitive business discussions into any commercial AI system effectively waives confidentiality. Companies should update vendor contracts and internal policies to explicitly prohibit inputting privileged information into external AI tools.