offaxis

5 ms·

1.
LLM Tool Failures: Only 3 Root Causes – Value, Condition, Intent(github.com)
7 points by offaxis 1mo ago|1 comments
2.
by offaxis 1mo ago
Reasoning is a strength in conversation, but it becomes a problem in execution. So instead of asking “Did you actually run it?” and only analyzing the outcome, the model needs to have its authority to decide input values taken away, leaving
3.
Forms did five things. Natural language kept one
3 points by offaxis 2mo ago|0 comments
4.
From knowledge-based inference to presence-based verification(discuss.huggingface.co)
2 points by offaxis 2mo ago|0 comments
5.
by offaxis 3mo ago
I think part of expertise is the checklist experts carry in their heads: What needs to be done? What counts as done? What must not be changed? When should the agent stop? What signals verify success? The important question is how that exper
6.
If Unsure, Ask. Never Guess. AI Agent Pre-Execution Checklist(discuss.huggingface.co)
2 points by offaxis 3mo ago|0 comments
7.
by offaxis 4mo ago
This is a really nice first game. The “zero asset files” constraint is especially interesting. Drawing the visuals in code, synthesizing the sound effects, and generating the music at runtime makes it feel like a real technical experiment,
8.
Unknown Is Not False: An AI Agent Pre-Execution Checklist(discuss.huggingface.co)
3 points by offaxis 4mo ago|0 comments
9.
by offaxis 4mo ago
I have been writing for the past few months since starting my new job, and this writing is a huge encouragement to me. It makes me realize once again that I need to practice consistently rather than blaming my lack of ability. I think I was
10.
by offaxis 4mo ago
I am still using GPT 5.5. Should I switch back to the Claude now?
11.
by offaxis 4mo ago
To be honest, I often feel that AI writes better than I do. Since AI drafts drafts quickly, my work speed has increased by about 80% compared to before.