Ethan He
speaker
722 appearances
1 recordings
1 series
first heard Jun 2026
last heard 1 Jun
Ethan He’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 1 in all, peaking in Jun 2026 with 1.
Appearances
And we've been trying to solve the context-length issues through various ways.
One thing is just brute forcing train longer context lengths.
Another is to manage the context better.
I think the same thing in language model is also going to be happening soon.
So for example, like the language models, they're not aware of how long their own context length is.
Once they hit like 80% or something, the automatic context compression is getting triggered.
And the model is not aware of that when it's working.
And maybe it's good for the models to know, oh, I'm approaching like 80% or something.
And something also pretty interesting, for example, in OpenClaw, every time you type in something, the current local time is automatically attached to your message.
So the model actually knows what time is it.
This is making the model timer there.
And also in tool calling, a lot of the intermediate tool call results automatically prune.
So there's context removal, context addition, and context compaction.
So all of these are
from the harnesses themselves.
From our experience, the heuristic engineering also have the models get absorbed into the models themselves.
I guess that's something very interesting to explore.
I'm imagining what if the model have access to the code of the agent harness itself and be able to modify it to whatever you want.
Say if the agent harness is short enough, you can just put it in the context in the system prompts.
And then the model will say, when I want to spawn a future version of myself, I can modify the agent harness.
Showing 661–680 of 722 · page 34 of 37
← Previous
Next →