Mmastodon TechnologySoftware first seen 1 d ago, last 1 d ago, peak #7
Sonnet 5.5 cost 14x more than Haiku with no gain
Original: Sonnet 5.5 cost 14x more per run than Haiku 5.5 and passed no more runs. What fixed it was telling the agent which helpe
A developer reports that Anthropic's Sonnet 5.5 cost 14 times more per run than Haiku 5.5 while passing no additional test runs in a coding task. The fix was not a bigger model: telling the agent which helper function in the repository was dead code resolved the problem. The post, written in the context of Claude Code workflows, argues that precise task instructions can matter more than paying for a costlier model, prompting discussion about model choice and prompt quality in AI-assisted programming.
Why now: Developers are weighing whether expensive frontier models are worth it when better instructions achieve the same results.
AnthropicSonnet 5.5Haiku 5.5Claude Code
Rank over time, top of the chart is #1. 2 snapshots from 1 d ago to 1 d ago.
Evidence
API: https://socialmediatrends-api.osmike.com/v1/trends/1877139