I’ve sent Claude back to look at the transcript file from before compaction. It was pretty bad at it but did eventually recover the prompt and solution from the jsonl file.
I did not know it was SQLite, thx for noting. That gives the idea to make an MCP server or Skill or classical script which can slurp those and make a PROMPTS.md or answer other questions via SQL. Will try that this week.
I was also disappointed by the lack of Jupyter notebooks support: I ended up not using Jupyter notebooks that much anymore, and when I do, well, I run them in Jupyter
For me, Safari sometimes randomly refuses to execute the search for the terms I entered: at that point I need to bring the search bar back up -> search terms are gone -> x -> bring search bar back up -> search terms are back there -> enter
I wish they stopped adding features, especially useless UI “improvements” and AI stuff nobody asks for, and focused on making the system rock solid as we’re used to.
I feel like it’s the opposite: the copy-paste issue is solvable, you just need to equip the model with the right tools and make sure they are trained on tasks where that’s unambiguously the right thing to do (for example, cases were copying code “by hand” would be extremely error prone -> leads to lower reward on average).
On the other hand, teaching the model to be unsure and ask questions, requires the training loop to break and bring a human input in, which appears more difficult to scale.
> On the other hand, teaching the model to be unsure and ask questions, requires the training loop to break and bring a human input in, which appears more difficult to scale.
The ironic thing to me is that the one thing they never seem to be willing to skip asking about is whether they should proceed with some fix that I just helped them identify. They seem extremely reluctant to actually ask about things they don't know about, but extremely eager to ask about whether they should do the things they already have decided they think are right!
> The extra typing clarification in python makes the code harder to read
It’s funny, because for me is quite the opposite: I find myself reading Python more easily when there are type annotations.
One caveat might be: for that to happen, I need to know that type checking is also in place, or else my brain dismissed annotations in that they could just be noise.
I guess this is why in Julia or Rust or C you have this stronger feeling that types are looking after you.
I think the face they fundamentally don't look after you is where my resistance comes from. Will try and evaluate some newer code that uses them and see how I get on a bit more :)
reply