210:13
Don't waste 2 years learning to build self-improving AI agents....
Hanako@hanakoxbt
87 views
Aug 07, 2026
~1 min read
Don't waste 2 years learning to build self-improving AI agents.
Stanford just dropped a 3.5 hour course on building them from scratch.
0:00 - self-improving agent basics
1:06:29 - multi-step reasoning in agents
2:20:08 - agents learning from feedback
look at the third one, because that is where the whole thing turns.
Attempt → Signal → Correction → Better attempt
an agent cannot improve from its own opinion of its work. the signal has to come from outside it, or the loop just reinforces whatever it already believed.
self-review is the old workflow. external correction is the new one.
that is the difference between a system that gets better every week and one that gets more confident every week.
Anthropic pays $750,000 a year to engineers who can build this.
bookmark and watch it today, then read the eval setup below
Stanford just dropped a 3.5 hour course on building them from scratch.
0:00 - self-improving agent basics
1:06:29 - multi-step reasoning in agents
2:20:08 - agents learning from feedback
look at the third one, because that is where the whole thing turns.
Attempt → Signal → Correction → Better attempt
an agent cannot improve from its own opinion of its work. the signal has to come from outside it, or the loop just reinforces whatever it already believed.
self-review is the old workflow. external correction is the new one.
that is the difference between a system that gets better every week and one that gets more confident every week.
Anthropic pays $750,000 a year to engineers who can build this.
bookmark and watch it today, then read the eval setup below
