Skip to content
editama.tech

AI Lab / Experiments in public

A public laboratory for agents and automation.

04Entries in the log

Each entry records what was tried, how it is built, what happened and what is still unknown, including the parts that did not work. Experiments move from exploring to active to archived, and the notes stay.

How experiments are tracked

Exploring

Early research. Questions outnumber answers.

Active

Running and maintained. Notes are updated as it changes.

Archived

Concluded. Kept for the record, with what was learned.

Category

Status

Lab log

4 of 4 entries

  • EXP-03

    Hermes Agent deployment

    Running Nous Research's open-source Hermes Agent as an always-on assistant on a small Ubuntu server, reachable from Telegram.

    ActiveSample
    Agent Infrastructure
    • Hermes Agent
    • Ubuntu Server 24.04
    • systemd
  • EXP-02

    Telegram AI agent workflows

    Using Telegram as the interface for small AI workflows, with a webhook that answers fast and a worker that does the slow, retryable part.

    ActiveSample
    Workflow Automation
    • Telegram Bot API
    • Python
    • SQLite
  • EXP-04

    Self-hosted AI memory systems

    Comparing ways to give a self-hosted agent memory that lasts: short working context, distilled long-term facts, and retrieval that mixes keywords with vectors.

    ExploringSample
    Self-hosted AI
    • SQLite
    • Embeddings
    • Python
  • EXP-01

    Autonomous agent loops

    A bounded plan, act, check loop for autonomous agents, run inside throwaway containers with hard limits on steps, time and spend.

    ArchivedSample
    AI Agents
    • Python
    • Docker
    • LLM API