L2 The agent loop
Fresh look each step
Rebuilds what the model sees from the saved record and a fresh look at the world before every step.
Why it matters
An agent working from an old copy acts on a screen, file or history that no longer exists. It repeats work, clicks things that are gone, or misses the user's interrupt.
What to build
Rebuild every request from the saved session
The default the research recommends. The reasoning and the code behind it are in the app.
What it takes, by kind of agent
- Conversational assistantMust-have
- CodingMust-have
- Deep researchMust-have
- Browser/GUIMust-have
How well it is studied
- Harnesses checked
- 30
- Checked facts behind it
- 253
See the 253 checked facts and how each harness does it
The facts, the code lines they come from and the rules for this topic are in the app. Sign in with Google; new accounts are approved by hand.
Also in L2 The agent loop
- Loop disciplineSettled practice
One plain loop runs the agent: call the model, run the tools it asks for, and repeat until it answers.
- Assistant Must-have
- Coding Must-have
- Research Must-have
- Browser Must-have
- BudgetsOpen problem
Caps on money, time, tokens and repeated actions that stop a run cleanly when one runs out.
- Assistant Must-have
- Coding Must-have
- Research Must-have
- Browser Must-have
- StoppingSettled practice
Ends a run in a clear, named way, hands back the work done, and checks that "done" really means done.
- Assistant Must-have
- Coding Must-have
- Research Must-have
- Browser Must-have