Agent Systems Lab: when review becomes the bottleneck
Four execution slots, one reviewer, two queues. A reproducible model of why faster execution can move waiting without removing it.
Four execution slots, one reviewer, two queues. A reproducible model of why faster execution can move waiting without removing it.
Reglas pequeñas, skills bajo demanda, decisiones fuera del chat y límites entre herramientas. Ejemplos de mi arnés personal, con enlaces al código.
Una vez releído en papel, “Menos software, más impacto” de Eduardo Ferro Aldama ocupa su sitio en mi mini-olimpo del desarrollo de software.
How we use AI matters when learning something new. A reading note on comprehension, debugging and the skills needed to supervise the work.
As implementation and coordination change, the interesting work moves toward framing problems, designing systems and making decisions.
Casi ocho años en Eventbrite, de Staff Engineer a Engineering Manager. Un agradecimiento a las personas y a una forma de construir equipos.
The early experiments: bringing intentional compaction into Cursor and keeping intent and context useful through longer coding sessions.