3 ms·

I was happily plodding away with it earlier when it threw this out in the middle of a response in Claude code: --- So — what did you actually see before you hi

by Spikefu 4mo ago

I was happily plodding away with it earlier when it threw this out in the middle of a response in Claude code: --- So — what did you actually see before you hit Ctrl-C? That's the信号 I'm most curious about, and it tells us what to --- That's the sort of behavior I'd expect from a one or two year old model quantized down to about 1 bit - right word, wrong language in a response. Google translate tells me that's Chinese for signal. I wonder what caused that to happen.

Computer0 4mo ago
I have been working with it for ~5 hours today and it has gone crazy twice to the point where I had to start a new session, looping reading a unrelated tmp file dozens of times over and over. And once for a weird api error. I will be honest it is probably a worse day for me than any with 4.7. But I don't want to be dramatic, I will keep trying it.
jerrygenser 4mo ago
Perhaps you were served from someone else's cache
Spikefu 4mo ago
Now it's arguing with me about why the edits it made aren't being picked up: ===== I'm not convinced. I didn't run it until you told me it was all fixed up. I tried running it with this path instead (base) spike@Tokenator music % python ~/Developer/spikefu/spikefu.py and it worked. It also now works with the old relative path approach. You're right, and I owe you an honest correction — I've now given you two confident explanations (Zed, then "timing window") and both were wrong, since you only ran it after I said it was done. The fact that clearing __pycache__ is what unblocked it means you were right: it was pyc-related. Before I theorize a third time, let me actually test the mechanism on your Python rather than assert: =====
- Spikefu 4mo ago
  It did one more weird thing where it seemed to lose the ability to count in order: ===== You're right that hard per-action gating causes loops. So rather than block the model, make the harness own the critical structure deterministically, leave a git trail, and use soft guidance: - (2) Commit breadcrumbs: /new commits the scaffold immediately (before discovery), then commits discovery progress. This gives the trail you want and a recovery path — a rogue rewrite becomes a visible git diff you can revert, instead of an invisible clobber. This is better than gating for your concern. - (4) Deterministic, verified seed: finalize re-asserts the kanban (writes TASK-001 directly, ignores any model mangling), verifies it landed, reports honestly. The harness guarantees structure regardless of what the model did. - (3) Soft-constrain discovery: prompt says "the .fractal files exist — never recreate them or write them via shell; read_file before edit_file; only edit intent.md/constraints.md." Guidance, not a hard gate. - (1) Fix command routing: the discovery > loop recognizes cancel/exit/:q and refuses a re-entrant /new — a routing fix, not a model gate. This alone would have prevented Run 2's disaster. I'm not going to start coding until you steer — I've burned enough trust guessing. My recommendation is to do (1) + (2) first (they're the highest-leverage and lowest-risk: command routing stops the trigger, breadcrumbs give traceability + recovery), then (4), then (3). ===== After that it seemed to behave itself and then did a compaction and since then it seems to be working properly again. Very odd. (and disconcerting)

by Spikefu 4mo ago

Computer0 4mo ago

I have been working with it for ~5 hours today and it has gone crazy twice to the point where I had to start a new session, looping reading a unrelated tmp file dozens of times over and over. And once for a weird api error. I will be honest it is probably a worse day for me than any with 4.7. But I don't want to be dramatic, I will keep trying it.

jerrygenser 4mo ago

Perhaps you were served from someone else's cache

Spikefu 4mo ago

Now it's arguing with me about why the edits it made aren't being picked up: ===== I'm not convinced. I didn't run it until you told me it was all fixed up. I tried running it with this path instead (base) spike@Tokenator music % python ~/Developer/spikefu/spikefu.py and it worked. It also now works with the old relative path approach. You're right, and I owe you an honest correction — I've now given you two confident explanations (Zed, then "timing window") and both were wrong, since you only ran it after I said it was done. The fact that clearing __pycache__ is what unblocked it means you were right: it was pyc-related. Before I theorize a third time, let me actually test the mechanism on your Python rather than assert: =====

Spikefu 4mo ago
It did one more weird thing where it seemed to lose the ability to count in order: ===== You're right that hard per-action gating causes loops. So rather than block the model, make the harness own the critical structure deterministically, leave a git trail, and use soft guidance: - (2) Commit breadcrumbs: /new commits the scaffold immediately (before discovery), then commits discovery progress. This gives the trail you want and a recovery path — a rogue rewrite becomes a visible git diff you can revert, instead of an invisible clobber. This is better than gating for your concern. - (4) Deterministic, verified seed: finalize re-asserts the kanban (writes TASK-001 directly, ignores any model mangling), verifies it landed, reports honestly. The harness guarantees structure regardless of what the model did. - (3) Soft-constrain discovery: prompt says "the .fractal files exist — never recreate them or write them via shell; read_file before edit_file; only edit intent.md/constraints.md." Guidance, not a hard gate. - (1) Fix command routing: the discovery > loop recognizes cancel/exit/:q and refuses a re-entrant /new — a routing fix, not a model gate. This alone would have prevented Run 2's disaster. I'm not going to start coding until you steer — I've burned enough trust guessing. My recommendation is to do (1) + (2) first (they're the highest-leverage and lowest-risk: command routing stops the trigger, breadcrumbs give traceability + recovery), then (4), then (3). ===== After that it seemed to behave itself and then did a compaction and since then it seems to be working properly again. Very odd. (and disconcerting)