sPTC latency sequence
Sequence diagram showing speculative sub-agent work starting during token streaming and later returning through a future store when the real REPL reaches the call.
root model
token stream
speculator
shadow repl
future store
promise ids
sub-agent
slow call
real repl
final exec
PARTIAL CODE
START EARLY
RESOLVE
FINAL CELL
GET FUTURE
RETURN NOW
the slow call runs while the root model is still writing
LEGEND
speculative work
return or final path