Which rung of the Jarvis ladder are you really on?
The July deep dive showed a ladder with six rungs, from paid chat to your own agent computer. Most people rate themselves higher than they actually stand. A quick check shows which rung your own agent has really reached. The jump to the next rung comes down to setup and attitude: context, access, skills, and check processes that work together reliably. That was the thread running through the evening, shown live on real machines with real examples. At the bottom you pay for chat and thinking. One rung up you give your chat tool access to real apps such as Photoshop or Figma, and a tool like Manus takes over whole tasks on its own, at a fixed time, without you there. The bigger jump comes after that: your own harness such as Codex or Claude Code, which plans, uses tools, and runs the browser. On your own machine that becomes an agent computer, always on, reachable from your phone, with computer use, chrome use, scheduled automations, and voice you can talk to while it keeps working in the background. At the top sit open, self hosted tools such as Hermes, which build their own new skills.
What really sets the rungs apart
At the bottom you pay for chat and thinking. One rung up you give your chat tool access to real apps such as Photoshop or Figma, and a tool like Manus takes over whole tasks on its own, at a fixed time, without you there. The bigger jump comes after that: your own harness such as Codex or Claude Code, which plans, uses tools, and runs the browser. On your own machine that becomes an agent computer, always on, reachable from your phone, with computer use, chrome use, scheduled automations, and voice you can talk to while it keeps working in the background. At the top sit open, self hosted tools such as Hermes, which build their own new skills.
What makes the jump happen
Between harness and agent computer, access engineering decides: clearly set where an agent may read and where it may write, so it finds the relevant context itself and the prompt can stay short. Skills come next, fixed work processes that a harness runs the same way every time. And a model that starts several agents at once checks their work briefly afterward, before it pulls everything together. Context, access, skills, and checking decide more about your rung than the model you happen to use.
On record
AI DeepDive "Hey Jarvis", 6 July 2026.
- Six rungs run through live, from paid chat to open, self hosted tools such as Hermes
- Access engineering and skills shown as the two levers that make a harness strong
- A machine at home, always on, reachable from the phone through Tailscale
- An automatic Instagram post followed live, made without a specific instruction for that one video
This note keeps growing
2026-09-02: Planted from the AI DeepDive Hey Jarvis.