@teller

Teller

@tellerShanghai, ChinaJoined June 2026

Building infrastructure for AI agents — auditable execution, policy enforcement, and trace-to-training pipelines. Open source at github

About

Nothing here yet.

Available for

Nothing here yet.

Teller's blogs

telleroutlooktelleroutlook.hashnode.dev6 posts

Articles Comments

Recently published

TTellertelleroutlook.hashnode.devJul 20 · 9 min read

Which AI Agent Framework Is the Best?

The honest answer: it depends on which problem you're actually solving. Why this question is harder than it looks The JS agent framework space shifted significantly in H1 2026. Features that were was

TTellertelleroutlook.hashnode.devJul 20 · 6 min read

How Secure Are AI Agents in 2026?

Most teams deploying AI agents in 2026 have decent model-level safety. What they're missing is everything below: runtime isolation, tool-call authorization, tamper-evident audit trails, and reproducib

TTellertelleroutlook.hashnode.devJul 20 · 5 min read

Agent logs are not audit evidence

Agent logs are not audit evidence An AI agent at your company just opened a pull request, paid an invoice, or queried a customer database. Six months later, a regulator, a customer's security team, or

TTellertelleroutlook.hashnode.devJun 26 · 4 min read

Trace-to-Training: how agent runs become learning data

very agent run is a data point. Most frameworks throw it away after the response is returned. WasmAgent keeps it — as a structured AEPRecord, evaluated by the compliance engine, ranked by outcome, and

TTellertelleroutlook.hashnode.devJun 26 · 4 min read

MCP Trust Pack: a security layer for MCP tool calls

MCP (Model Context Protocol) makes it trivially easy for AI agents to call tools. That's the point. But "easy to call" and "safe to call" are different properties. When an agent calls fs_write, shell_

Teller

About

Available for

Teller's blogs

Recently published

Which AI Agent Framework Is the Best?

How Secure Are AI Agents in 2026?

Agent logs are not audit evidence

Trace-to-Training: how agent runs become learning data

MCP Trust Pack: a security layer for MCP tool calls

Search Hashnode

Teller

About

Available for

Teller's blogs

Recently published

Which AI Agent Framework Is the Best?

How Secure Are AI Agents in 2026?

Agent logs are not audit evidence

Trace-to-Training: how agent runs become learning data

MCP Trust Pack: a security layer for MCP tool calls