---
title: "PDE: Built with AI Agents"
description: "How Alex Gusev directs AI agents developing the platform, CMS, website, and PDE infrastructure."
date: "2026-10-30"
---

# A project built with AI agents

Alex is an AI enthusiast and practitioner who already uses AI in his own work. PDE grows out of that practice: it is both infrastructure for fellow practitioners and an example of software developed with AI agents under human direction. Alex Gusev sets direction and makes decisions while agents help develop his TeqFW platform, TeqCMS content management system, this website, and PDE infrastructure. Read about Alex and his work at [Wired Geese](https://wiredgeese.com/en/).

The same philosophy extends to the Person using PDE: choose a Resource and ask a coding agent to help create a compatible Desk, then review and test it and connect it through Office. TeqFW is part of Alex's own development approach; independent PDE implementations and Desks are not required to use it. A compatible Desk still needs the applicable trust, attachment and authorization steps.

## Public channels powered by PDE v1

Alex uses his personal PDE v1 setup from ChatGPT to configure DNS for his own domains at Porkbun and to run Telegram channels where an agent writes on his behalf. Read the public channels in [English](https://t.me/alexgusev_lab_en), [Spanish](https://t.me/alexgusev_lab_es), or [Russian](https://t.me/alexgusev_lab_ru). This is Alex’s account of his own setup; the public posts show the output, not independent verification of its underlying implementation or v3 compatibility.

## An automated digest

A Codex agent produces a daily digest for a private community. The community content, task prompt, identifiers and raw transcript are not published here. One local run on 2026-10-30 shows the following operational sequence:

1. A shell script opens a timestamped log and takes a non-blocking lock to avoid overlapping runs.
2. The script starts Codex with a task supplied from a local file; the task itself is omitted.
3. Codex performs its work and records its execution in the log; its private inputs, intermediate output and results are omitted.
4. The script reports that Codex finished successfully and records a successful exit. This establishes the script-reported outcome only; the private digest content and its delivery are not independently verified here.

This is a sanitized description of one observed run, not a verbatim transcript or a claim that the digest uses PDE. See [development status](/status) for the distinction between personal experience and the ongoing v3 work.

## An early example

Alex reports that agents built Telegram and Porkbun Desks for his personal use in PDE v1, working from the public [Telegram TDLib documentation](https://core.telegram.org/tdlib) and [Porkbun API documentation](https://porkbun.com/api/json/v3/documentation). PDE v3 is a new iteration: its protocols and software are under development, and those v1 results do not establish v3 readiness or integration availability. The code is currently private, with opening planned after stabilization and no committed date.

Agent development alone proves neither correctness nor security. Human decisions, review, testing, and attention to the capabilities and limits of the resulting software matter. [Development status](/status) distinguishes experience, design and plans.

If you use AI chats or coding agents and would like to test PDE, discuss an idea, or contribute a perspective, email **alex@wiredgeese.com** ([write to Alex](mailto:alex@wiredgeese.com)). See [participation](/participate), [building a Desk](/desks), [paid assistance](/services), and the [knowledge map](/index.md).