By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
At a glance
WHAT IT’S REALLY ABOUT
Weave Router automates model choice to optimize AI coding cost
- Weave is described as a system that measures engineering productivity and quality, then uses those measurements to optimize how humans and coding agents deliver software.
- Weave Router routes each coding prompt/task to the best-fitting model automatically, aiming to deliver comparable results with lower cost and faster speed than always using a top-tier model.
- The founders emphasize that effective AI coding requires measuring more than outputs, including downstream quality signals such as bugs and longer-term production impacts.
- Teams’ needs are presented as evolving from AI adoption, to “token maxing,” to optimization—covering model choice, prompting, agent parallelization, and avoiding wasteful loops.
- The episode frames the future of engineering management as “token allocation,” where high performers combine taste and direction with effective orchestration of AI agents.
IDEAS WORTH REMEMBERING
5 ideasWeave positions itself as measurement infrastructure for human-and-agent engineering.
He says Weave builds models to measure engineering work and uses that measurement to optimize how humans and coding agents work together, with the end goal of making a company operate like a “software factory.”
Weave Router aims to pick the right model per task automatically.
He describes Router as a component you plug into a coding agent so each prompt/task is automatically routed to the “best model,” removing the need for developers to manually choose between models.
Routing is framed as a way to increase output while reducing cost and latency.
He says benchmarks show they can reach “Astra-level performance” at “half the cost and twice the speed,” and later claims that for users already maxing paid plans, they’ll “go like twice as far” for the same spend by using smaller/cheaper models when appropriate.
Quality is treated as a first-class constraint, not just token savings.
He argues routing can be safe because Weave measures both velocity and quality, and cites Weave’s internal use: output continues to accelerate with no increase in bugs/quality issues, and the team sits in the “top 1%” in their own data.
Weave Router is designed to optimize within real-world subscription and subsidy constraints.
He explains Router can route across multiple subscriptions/accounts to maximize usage of subsidized plans first, then “sprinkle in” open-source models when they’re best even considering subsidies.
WORDS WORTH SAVING
5 quotesWeave is a tool to turn any software company into a software factory.
— Adam Cohen
What Weave Router does is it sends every task to the best model for that task.
— Andrew Churchill
You don't always need a bazooka where a handgun will do.
— Andrew Churchill
One of the main metrics we show is like cost per output.
— Adam Cohen
Running an engineering team will be an exercise in token allocation.
— Andrew Churchill
SPEAKERS
Adam Cohen
guestCo-founder of Weave; discusses engineering productivity metrics, agent evaluation, and model-routing economics.
Andrew Churchill
guestCo-founder of Weave; focuses on the Weave Router’s technical approach to routing coding tasks across LLMs and evaluating outcomes.
High quality AI-generated summary created from speaker-labeled transcript.
