Skip to content

Standard Reader

Jeff Auriemma

Jeff Auriemma

@jeff.auriemma.xyz

APIs in the age of AI. Talking Open source, GraphQL, MCP, and more

0Readers

12Posts

6d agoLast active

Open source token efficiency benchmark update, inference taxonomy framework, Alec Aivazis on Houdini and GraphQL

Week of 2026-09-07 roundup

A functional taxonomy for LLM inference in agentic tasks

A framework for understanding where LLM inference goes during agentic tasks: Initialization, Reasoning, Orchestration, and Synthesis. Consider "Reasoning Yield" as a way to measure how much inference is spent resolving uncertainty rather than mechanically advancing already-determined execution.

MCPJam's Prathmesh Patel, another bad GraphQL take, AI WG, and conference preview

Week of 2026-08-31 roundup

MCP + GraphQL, Robert Balicki on Isograph, HTTP QUERY, GraphQL clients

Week of 2026-08-24 roundup

A2A: Adversary to Adversary?

When do you choose Agent2Agent over MCP or subagents? It all has to do with administrative boundaries.

a2aagentic development

Future of staff-level engineering, MCP auth, CLI dev, new GraphQL tagline, governance by default, API vs. Tool design

Week of 2026-08-10 roundup

engineering managementmcp

GraphQL Codegen, stateless MCP, prompt-as-policy, persisted queries

Week of 2026-08-03 roundup

Stateless MCP is cheaper to scale

MCP’s 2026-07-28 specification removes protocol sessions and the `Mcp-Session-Id` header, making remote MCP servers cheaper and easier to scale. This post explains the old session model, why it required infrastructure such as sticky routing and Redis, and how stateless MCP preserves application state through explicit handles.

mcpsoftware development

You can't prompt your way into policy

Putting policy inside plain-language instructions gives the model a lot of (expensive) work to do before it can act, and there are no guarantees about safety. Furthermore, other prompts in the context window may be pushing in a different direction. Here's a vision for a better governance model via a context graph.

mcpinformation security

Hype cycle, bad takes, API design, GraphQL + AI, and marketing

graphqlsoftware development

Context graphs, MCP tools, plain text, making users awesome, and events!

MCP tool calling under the hood

So when we say that an AI called a tool, what usually happened is a little more specific than that phrase suggests. Here's a deeper look into that plumbing.

mcpsoftware development