L1 Context model
Message model
One message format for the whole conversation, in which every tool call stays matched to its result.
Why it matters
Model providers reject a conversation in which a tool call has lost its result, and the error shows up far from its cause. One format with matching built in prevents that and keeps a saved and reloaded conversation identical to the original.
What to build
One record per tool call, repaired at render
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
- 257
See the 257 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.