pholden

At least in one of these cases the models were being tested without the normal safeguards you would see as a regular user.

> A subagent investigates a bug, turns its findings into a nice human-readable summary, the parent agent reads that summary, and then turns it into another nice human-readable summary for you.

Is that a problem with https://code.claude.com/docs/en/output-styles?

> Output styles apply to the main conversation only: a subagent runs its own system prompt, so styles don’t change how subagents respond. A fork is the exception, because it inherits the parent’s full system prompt.

Hi - I wrote n64js.

Yes, it's true that Super Mario 64 is one of the easiest roms to get running as it uses relatively few features.

The format (z64 vs v64) shouldn't matter - I byteswap the file on load. I don't handle compressed roms though.

A few other roms that should boot: mario kart, wetrix, aeroguage, extreme-g, pilot wings, star fox. I haven't tested many others yet - let me know if you have success with anything else!

One thing to bear in mind: reload the page when you switch roms. There's some state I'm not resetting properly which breaks everything quite horribly. Sorry for the lameness - I'll try to get this fixed ASAP.

If you open the debug pane, you should get some debug output which gives you some clue as to why the rom failed to boot (usually unhandled memory accesses, unimplemented save format, unimplemented instruction, something like that). For the games that do boot, there are still some cpu bugs which prevent them from passing the CIC-chip's CRC check. E.g. GoldenEye and Zelda OoT both seem to fail in this way.