you can skip most agent content on this site if you read these seven:
1. ReAct by
@ShunyuYao12 et al (reasoning and acting in one loop, every agent framework is basically this)
2. Reflexion by
@noahshinn024 et al (agents that learn from their own failures without retraining)
3. Tree of Thoughts by
@ShunyuYao12 et al (what happens when you let it branch instead of guess once)
4. Generative Agents by
@joon_s_pk et al (the smallville paper, memory and reflection)
5. MemGPT by
@charlespacker et al (context windows as an operating system problem)
6. SWE-bench by
@ofirpress et al (the benchmark that made agent evals real)
7. Voyager by
@DrJimFan et al (an agent that writes its own skill library)
save this, add these to your weekend reading list