Coming soon
Athenaeum: a library of work for agents
Athenaeum is Vidbyte's coming library of real, valuable work for AI agents. Each entry, a volume, names a job worth doing and teaches your agent how to do it at an expert level, including how long to keep going and how to prove it is done. Volumes are designed to run persistently, proactively, and autonomously inside Claude Code or Codex, on your own machine and accounts. Athenaeum is not released yet: this page explains the concepts so you know what is coming.
What Athenaeum is
Agents can reason and use tools, but every session starts from a blank prompt that has to guess which work is worth doing and how an expert would do it. Athenaeum is a curated collection that answers both questions, one job at a time.
Anatomy of a volume
A volume is one packaged job. It is filed like a book, with a call number and a shelf, and it holds four parts.
The work — A real job worth an agent's time, defined by the result it produces for you, such as a ranked list of prospects or a codebase with a bug pattern fixed everywhere.
The method — Expert-level instructions: the steps, judgment calls, and hard-won rules a specialist follows, written so your agent does the job the way a professional would.
The run — How the work keeps going: once, until a goal is met, on a schedule, or whenever something changes. A volume remembers what it did last time.
The check — A clear test of done. A fresh agent that did not do the work checks the result before anything comes back to you.
A volume also lists what it needs from you before it starts: its inputs, the tools and accounts it will use, and its limits. You can read exactly what a volume will do before it does anything.
Persistent, proactive, and autonomous runs
Real work rarely fits in one conversation, so volumes are designed to keep going.
Persistent — Keeps working on a job across runs and remembers what it already found, so later runs build on earlier ones.
Proactive — Notices when there is something to do, such as a new competitor launch, and starts on its own.
Autonomous — Carries the job through its steps to a checked result without you steering every turn.
Many at once — One person can direct hundreds or thousands of agents, each working one volume and each checked before it reports back.
Where volumes run
Athenaeum gives your existing agent better work instead of replacing it. Volumes are designed for Claude Code and Codex on your own machine, so the model work stays on the account you already pay for. vidbyte-cli, the same CLI that runs Vidbyte's agents today, will run volumes, record each run, and hand each result to a fresh agent to check.
- Says what to do
- Agent skillYes, once
- Athenaeum volumeTeaches a tool or a workflow
- Names a job worth doing
- Expert method
- Agent skillWhatever you write
- Athenaeum volumeFor the tool it covers
- Built in, from a specialist's process
- Keeps running
- Agent skillNo
- Athenaeum volumeNo
- Persistent, proactive, autonomous
- Checks the result
- Agent skillNo
- Athenaeum volumeSometimes
- A fresh agent checks the test of done
| Prompt | Agent skill | Athenaeum volume | |
|---|---|---|---|
| Says what to do | Yes, once | Teaches a tool or a workflow | Names a job worth doing |
| Expert method | Whatever you write | For the tool it covers | Built in, from a specialist's process |
| Keeps running | No | No | Persistent, proactive, autonomous |
| Checks the result | No | Sometimes | A fresh agent checks the test of done |
The first shelves
Volumes are grouped into shelves by the part of your work or life they improve. The first planned shelves are Growth, Research, Engineering, Sales and outreach, Operations, and Life. A volume joins the library only after it has been run on real work and its result has been checked.
Next steps
Athenaeum builds on agents you can already run from vidbyte-cli: Persistent Researcher, the Suggestion Agent, Persistent Agent, Ensemble Agent, Task Board, and Stages Agent.