[dead]
[flagged]
Like the idea of distributed SQLite, but do we have benchmark? What use cases it has advantage?
DSA?
I feel many models are over fitting. During software development only opus can save my time. Sol needs some back and force. Other models such as muse wonder and argon are totally waste time need too much effort to make…
Yes I am also not happy with those agent terminals. Will try yours.
Like the idea. Did you try use AI to control it and do some intreating things?
Nice work, will you develope an IOS one?
will try and see how much I get.
Define the task, delegate, then review is pretty much how I work with agents now too. How does a card move to review: does the agent decide it's done, or do tests in the sandbox have to pass first?
I review a lot of AI-written code, so leaving comments on lines and having the agent read them through the CLI is the part I like most. Where are the comments stored? Do they stay with the branch?
Ha, yes. "It crashed" is not a proof. A test that fails before the fix and passes after would be a good bar.
Tried it but couldn't find an opponent. I'll try again tomorrow.
Yes. If you have a model scan for errors and fix them in a loop, it can go 100 rounds without stopping. Maybe it needs a strict stopping criterion, e.g. the model has to prove a vulnerability is actually exploitable…
I do it this way too. I think having the same lab's model review the code in a fresh session also helps, so I use both: I generally start a new Claude Code and a Codex and have them review the code with a focus on 3…
[dead]
[flagged]
Like the idea of distributed SQLite, but do we have benchmark? What use cases it has advantage?
DSA?
I feel many models are over fitting. During software development only opus can save my time. Sol needs some back and force. Other models such as muse wonder and argon are totally waste time need too much effort to make…
Yes I am also not happy with those agent terminals. Will try yours.
Like the idea. Did you try use AI to control it and do some intreating things?
Nice work, will you develope an IOS one?
will try and see how much I get.
[flagged]
[flagged]
[dead]
Define the task, delegate, then review is pretty much how I work with agents now too. How does a card move to review: does the agent decide it's done, or do tests in the sandbox have to pass first?
[flagged]
I review a lot of AI-written code, so leaving comments on lines and having the agent read them through the CLI is the part I like most. Where are the comments stored? Do they stay with the branch?
Ha, yes. "It crashed" is not a proof. A test that fails before the fix and passes after would be a good bar.
Tried it but couldn't find an opponent. I'll try again tomorrow.
Yes. If you have a model scan for errors and fix them in a loop, it can go 100 rounds without stopping. Maybe it needs a strict stopping criterion, e.g. the model has to prove a vulnerability is actually exploitable…
I do it this way too. I think having the same lab's model review the code in a fresh session also helps, so I use both: I generally start a new Claude Code and a Codex and have them review the code with a focus on 3…