Skip to content

Andy Triedman

Theory Ventures

  • 1 episode on Chain of Thought
  • 6 topics covered
  • 1 full transcript
  • Oct 2026 first on the show
Andy Triedman

Andy Triedman is a partner at Theory Ventures, where he works with AI-native companies building AI infrastructure and application platforms, and he is building his own software factory while watching how startups run theirs. On Chain of Thought he argued that most agent work is limited by context, not intelligence, and walked through which tasks still need human approval, when cheaper models are enough, and how to organize agent memory.

Cover art for For AI Agent Fleets, the Model Is Rarely the Bottleneck | Theory Ventures' Andy Triedman

Latest episode with Andy Triedman · EP 79

For AI Agent Fleets, the Model Is Rarely the Bottleneck | Theory Ventures' Andy Triedman

Andy Triedman, Theory Ventures ·

Episodes

What the conversations cover

These notes summarize the episodes featuring Andy Triedman, including other speakers in panel conversations. Follow a source link to read the discussion in context.

EP 79 For AI Agent Fleets, the Model Is Rarely the Bottleneck | Theory Ventures' Andy Triedman

  • Most agent work is context limited, not intelligence limited. Andy Triedman, a partner at Theory Ventures, says the models are “frankly all good enough” for most day-to-day tasks, and that as intelligence becomes a commodity, the value moves to the applications that deliver the right context to the right model at the right time. Episode takeaways
  • The biggest change Triedman sees is teams moving from solving one problem at a time with an agent to building systems that solve whole classes of problems. Instead of asking an agent for a CrowdStrike integration and then a Splunk connector, the companies with the most leverage ask for a system that develops and tests integrations, then hand it 20 new ones at a time. Episode takeaways
  • A one-paragraph feature spec plus the three or four follow-up questions a coding agent asks is “one level too high.” Triedman goes another eight to twelve questions deeper, so the agent isn’t left making assumptions that change the customer experience. In his own software factory, a product requirement might have a 20 to 30 question Q&A, which must be exhausted before engineering tickets are specified. Episode takeaways

In their words

I think unlike my … portfolio companies I work with, I'm probably a little more like laissez-faire and willing to like push the gas on the agent automation, which gives me a good sense of where it starts to break and where it's really strong.

Andy Triedman For AI Agent Fleets, the Model Is Rarely the Bottleneck | Theory Ventures' Andy Triedman · 13:33

Transcript excerpts; ellipses mark omitted words. Some automatically selected quotes have filler words removed.