
Decision Models Could Make Coding Agents Cheaper
Agent harnesses could reserve frontier models for substantive reasoning while cheaper decision models handle routine navigation, routing and tool selection.
Read the snack ↗
IntelligenceSnacksA new AI model can combine genuine improvement with a flattering comparison against an older model that seemed to deteriorate over time. Jev proved most useful when it ranked a steady stream of varied SEO signals for deeper analysis, rather than adding another scoring layer to a low-volume workflow.
With Pete Winn and Andy David
7 ideas developed from this conversation
About this conversation
A new AI model can combine genuine improvement with a flattering comparison against an older model that seemed to deteriorate over time. Jev proved most useful when it ranked a steady stream of varied SEO signals for deeper analysis, rather than adding another scoring layer to a low-volume workflow.

Agent harnesses could reserve frontier models for substantive reasoning while cheaper decision models handle routine navigation, routing and tool selection.
Read the snack ↗
Fast, structured model judgements can handle uncertain decision points while ordinary code remains in control of the wider automation.
Read the snack ↗
Jev proved most useful when it ranked a steady stream of varied SEO signals for deeper analysis, rather than adding another scoring layer to a low-volume workflow.
Read the snack ↗
Automating both ends of email can save individual effort while leaving the medium’s organisational burden intact.
Read the snack ↗
A new AI model can combine genuine improvement with a flattering comparison against an older model that seemed to deteriorate over time.
Read the snack ↗
Fast, specialised models can make sorting, routing and checking cheap enough to place throughout automated workflows.
Read the snack ↗
Persistent personal agents could weaken businesses that profit when customers postpone the work of finding a better deal.
Read the snack ↗