Masad: Purge fixed bug history from AI agent context memory
Amjad Masad · What Agents That Actually Work Look Like Right Now with Replit's CEO and Founder · Jun 9, 2026 · at 9:02
Replit CEO Amjad Masad outlines best practices for managing persistent memory and context compaction in autonomous developer agents.
0:00 / 0:15exact quote · 15.1s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“Bugs that it fixed should be removed from context because then it will get confused, but Architectural decisions on how it built things in the past is very important to stay in context or stay in long-term memory and be able to pull it in very easily.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Amjad Masad
Masad: Only software agents work; sales and marketing agents do not
“The only agents that work are software agents, marketing agents, sales agents, all that haven't really worked very well.”
Masad: Replit users were six months ahead of AI labs on agents
“In reality, actually, Reply users were like six months ahead of everyone else, even people in the labs, because when we launch these agents, a lot of times we get researchers messaging us, it's like, oh, we didn't know Claude was capable of running for, you kn…”
Masad: Many software engineers are laggards in adopting AI agents
“I actually think engineers are laggards. Not all engineers, but like a lot of engineers are because they're kind of tied to their workflows.”
Masad: AI context windows now exceed human context capacity
“I think we've already crossed a threshold. Yeah. For which the context links Is larger than any context that human can have. Right. And I think that's huge because, I mean, just think back two years ago we had 16 K context link. Right now we have a one million…”