Every AI Company Is Accidentally Building a Bank — Dor Sasson, Stigg

Agent products can overrun budgets when usage is checked only after invoicing. A reserve-before-inference, settle-after pattern with spend caps per team, user and agent gives builders a concrete design for cost control.

• Sasson cites the April pricing scramble: Anthropic cut third-party agents such as OpenClaw off subscription plans, OpenAI changed its Pro tier, and GitHub pulled free Copilot access.
• His diagnosis is that usage gets checked after the invoice; he proposes an ATM-style flow, a synchronous entitlement check before inference and asynchronous settlement after, and points to OpenAI's published architecture as an example.
• Once checks move up front, he says, the hard problems become concurrency and double spend, and deciding which credit pool gets drawn down first.

• inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
• AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.

• videoFrom 36% to 100%: How Self-Improving Agents Write Their Own Skills — Rafal Wilinski, Runlayer
• videoIs Speculative Decoding Worth It? Profiling vLLM on NVIDIA Blackwell — Akamai
• videoGrep or Embeddings? Agentic Search Over Company Documents — George He, LlamaIndex
• videoThe 6 Pillars of an Agentic Harness for Production — Varun Krovvidi, Resolve AI

• videoWhy Would AI Companies Want to Slow Down?a16z
• videoAI, Infrastructure, and the Next Investment Cyclea16z
• videoNotion's Token Town — Sarah Sachs, NotionAI Engineer

Checking sign-in…

Loading comments…