it took me two months but i finally got it working! the full Cardinal synth port with over 1000 modules. it almost broke me but it's finally working...
the interface still needs a lot of work and polyphony is limited. cardinal uses up most of the cpu, but you can get away with 4 voice polyphony.
this is not out yet, need to fix a few more things before i release it.
Hi all,
I'm trying to figure out what the difference is between the tape and disc modes, the manual says tape is for live looping and overdubbing but disc can do that too.
Tape support sos recording and so does disc, they can both overdub when playing in reverse.
Disc has the option to save the sample but Tape does not.
Did Disc just get all the features over time? The tape mode doesn't even have tape effects, so I'm not sure what the point is? Am I missing something?
I'm using the latest firmware, 2.2.0.
Thanks
The slice problem is when the agent implements "one small slice" of the task, not the whole thing.
This happens with plan mode, custom plans, adrs, simple asks like "do this", etc. It has happened with every model since 5.3-codex and every reasoning effort, medium, high and xhigh.
It's not very obvious for simple things but more obvious for harder tasks.
The layer problem is when agents need to update existing code, they add new subsystems, duplicate code and tests instead of updating existing code, even if it's been asked explicitly to read the lines.
So you might get a second layout system when asking for a new button, or you might get a new game shell that only proves tests.
This also happened with all models from 5.3 to 5.5 with every reasoning effort, medium, high and xhigh.
I am not new to codex or programming, i've burnt 100 billion tokens so far in the last 4 months, trying to figure out how to use this thing.
I am trying to build a game and a complex data pipeline project. I've also spent a significant amount of time trying to write my own skills/workflows.
I tried a lot of things, i rewrote my workflow/skill stack at least 5 times, i tried overriding the system prompt, i've watched every hype youtube video and checked out all the skills that are popular and none of them solve these issues.
How do you deal with these issues? what strategies or skills do you use to help with these? i get decent results if i review every single step and line of code and direct the agent manually, but as soon as i do "implement plan" or goal, i get slop.