The Twenty Minute VCWhy Token Maxing is Failing Enterprise Startups | Legora CTO
At a glance
WHAT IT’S REALLY ABOUT
AI makes code cheap; enterprises must optimize product and review workflows
- AI tooling has made writing code dramatically cheaper, shifting the main bottlenecks to product definition/PM throughput and code review quality rather than code creation.
- AI code review and multi-agent workflows exist today but remain immature, creating a pressing need for new review paradigms focused on architecture, security boundaries, and system-level impact.
- Legora reports that AI agents produce over half of their code, which accelerates shipping but increases security risk and raises the bar for guardrails, human oversight, and scalable process design.
- PMs can now prototype directly with AI to reduce handoff costs, but in companies like Legora the highest-leverage use of PM time remains customer discovery and synthesis because product work becomes the limiting factor.
- Enterprises should avoid incentivizing raw token usage (“token maxing”) and instead reward outcomes via demos and measurable output, while investing heavily in developer experience and agent-enablement infrastructure.
IDEAS WORTH REMEMBERING
5 ideasTreat AI tooling spend as opportunity-cost optimization, not a budget line item.
Lauritzen argues the cost of not gaining speed is often higher than token/tooling costs in competitive markets, so spend should be justified by velocity and impact rather than minimized by default.
The new bottleneck is product clarity and review, not typing code.
With code generation accelerating, teams must invest in PM efficiency, clearer handoffs, and better review mechanisms to prevent throughput from stalling at definition and merge time.
AI code review needs to evolve from line-by-line nitpicking to system-level evaluation.
He wants review to focus on architecture direction, stability, security boundaries, and strategic tradeoffs—areas where humans add leverage and where today’s tools are still weak.
Create guardrails so agents can move fast without breaking invariants.
As codebases and agent count grow, mechanistic enforcement (rules, constraints, approved data paths) becomes essential so agents can operate autonomously within safe boundaries.
Developer Experience becomes a force multiplier—especially when each engineer is “10x’d” by AI.
Legora’s DevEx team improves local setup, builds internal coding/review agents, and streamlines onboarding; Lauritzen regrets not staffing this earlier because small percentage gains compound across a highly-leveraged team.
WORDS WORTH SAVING
5 quotesThat is now super cheap, so that's sort of been compressed. And so the, the bottleneck now is, like, the, the two other ends, which is, uh, review. How can we do that much more efficiently? Um, and then it's, how can we actually do the product piece much more efficiently?
— Jacob Lauritzen
The job of a sys- of a, of an engineer is changing from typing a bunch of code to sort of one layer above it, which is, um, what does the system look like?
— Jacob Lauritzen
It's like, "This is who I am. This is who we are, a- and, and some of you are gonna hate it, and that's okay," um, because you need to have some edges.
— Jacob Lauritzen
Get, get a leaderboard, um, and, and bring up token usage at performance reviews, uh, and that leads to Token Maxing, which is people just burn tokens just to look good. Um, that's a really stupid way to do anything.
— Jacob Lauritzen
Honestly, just work harder than the 800-pound gorilla. People underestimate this. Like, the 800-pound gorilla, no one in the 800-pound gorilla is extremely excited to be there.
— Jacob Lauritzen
High quality AI-generated summary created from speaker-labeled transcript.
