More Effort, Or A Bigger Model?
Bad result? Diagnose before you rage-quit to another tool. Skipped files, no self-check, quit halfway — that's low effort: raise the effort level. Read everything and still confidently wrong — that's the model's ceiling: escalate the model. This one question settles 95% of 'the AI got dumber' complaints.
Why it works
Most 'the AI got dumber' complaints dissolve under one diagnostic question: did it lack knowledge, or lack diligence? Skipped files, no self-check, quit halfway — that's diligence: raise the effort setting. Read everything and still confidently wrong — that's capability: escalate the model. Fixing the wrong axis wastes money in one direction and patience in the other.
How to do it
- Bad result? Don't switch tools yet — autopsy it with the transcript.
- Check what it actually read: did it open the relevant files, or guess past them?
- Check what it verified: did it run/preview/test its own output, or declare victory blind?
- Missed inputs or skipped checks → raise the effort level next to the model picker.
- Full inputs, real checks, still wrong → that's the model's ceiling: escalate the model.
- Set effort once for your usual work rhythm; retune per task only for the outliers.
Copy this
prompt
Before I judge this result: list which files you read for it, what you verified after producing it, and what you skipped. Be honest — skipped is fine, hidden isn't.
What to watch for
- Raising BOTH knobs at once on every failure teaches you nothing and doubles the cost — change one axis, observe, then decide.
- Max effort on trivial tasks quietly multiplies token spend; effort controls how thoroughly it works, and thorough has a price.
Sources
Original: conon.ai