zomglings

Good game, but the fact that there is a target that you have to achieve in a fixed number of rounds changes the math.

For instance, if all I care about is satisfying your win conditions, and I am above $20,000, then I should just pass till the end. Similarly, if I have been unlucky (even with Kelly bets) for the first 14 rounds, I should go all in on the 15th round.

Could you share the restaurant? I maintain a list.

God I miss that show.

It doesn’t have to be $300/day worth of output tokens. It could be like $290/day worth of input tokens to teach both you and the model about the problem you are solving and then $10/day worth of output tokens.

I have been hungry to do more work from my cell phone. It's ridiculous to be forced to sit in front of a computer to work with AI.

My current solution is to have claude (--dangerously-skip-permissions) listen for messages in my slack DMs to myself and take action in response to those messages.

I would happily switch to something better.

Why is Omnara better?

Honestly have never heard of this person. Why does HN care so much that he joined OpenAI?

Does anyone else find the use of different shades of green for the graph comparing Gemini 2.5 Pro and Sonnet just a little insane?

They did it!

Yes, this is true. It's a summary, and cannot really undo a /clear. It is just a directed, cheaper /compact.

Apologies for the late reply. My kids demanded my attention yesterday.

It doesn't seem to have included any points on style or workflow in the context. Most of my context documents end up including the following information:

1. I want the agent to treat git commits as checkpoints so that we can revert really silly changes it makes.

2. I want it to keep on running build/tests on the code to be sure it isn't just going completely off the rails.

3. I want it to refrain from adding low signal comments to the code. And not use emojis.

4. I want it to be honest in its dealings with me.

It goes on a bit from there. I suspect the reason that the models end up including that information in the context documents they dump in our sessions is that I give them such strong (and strongly worded) feedback on these topics.

As an alternative, I wonder what would happen if you just told it what was missing from the context and asked it to re-dump the context to file.

Fair enough. That's pretty cool, I haven't gone that far in my own work with AI yet, but now I am inspired to try.

The point is to get a pipeline working, cost can be optimized down after.

What is in the file?

You can ask it to store its current context to a file, review the file, ask it to emphasize or de-emphasize things based on your review, and then use `/clear`.

Then, you can edit the file at your leisure if you want to.

And when you want to load that context back in, ask it to read the file.

Works better than `/compact`, and is a lot cheaper.

Thanks, makes sense that an MCP server that edits files is a workaround to the problem.