Why Token Maxing Is Failing Enterprise Startups | Legora CTO

The Twenty Minute VC (20VC)
The Twenty Minute VC (20VC)Jun 6, 2026

Why It Matters

AI tooling is turning code writing into a commodity, forcing enterprises to redesign engineering roles, product workflows, and security safeguards to stay competitive.

Key Takeaways

  • AI tooling slashes developer code‑writing time, shifting bottlenecks to review.
  • Legora uses both Cursor and Cloud Code, achieving ~2% AI‑generated code.
  • Future engineering will focus on system design, not line‑by‑line coding.
  • AI‑driven code review and guardrails are critical for security at scale.
  • Product teams can prototype instantly, front‑loading value before engineering involvement.

Summary

The interview with Legora CTO Jacob Lorettson centers on why traditional token‑maxing models are breaking down for enterprise startups and how AI‑driven tooling is redefining software development. Lorettson explains that the company’s rapid growth to $100 million ARR in 18 months is powered by AI assistants such as Cursor and Cloud Code, which let engineers produce far more code than before, compressing the historic “write‑code” bottleneck.

With code generation now cheap, the new constraints lie in code review, product definition, and system‑level architecture. Legora runs AI‑based review bots for security and design checks, and the CTO stresses that future engineers will spend most of their time shaping system topology and guiding autonomous agents rather than typing individual lines. He also notes that only about 2 % of the codebase originates from AI, but the agents are already outperforming human contributors.

Specific examples illustrate the shift: AI agents negotiate guardrails, automatically generate post‑mortems, and even run incident‑response loops without human wake‑ups. Lorettson warns that the surge in AI‑generated code raises novel security threats, prompting continuous human PR reviews and the development of risk‑scoring mechanisms. He highlights the importance of meta‑engineering teams that tune agents, collect data, and enforce policies.

The broader implication is a restructuring of engineering organizations. Product managers can prototype and validate ideas at speed, reducing the need for early‑stage engineering effort, while senior engineers move up to oversee architecture, agent effectiveness, and security guardrails. Enterprises that adopt these practices stand to accelerate delivery, but must invest in new review frameworks and talent to manage AI‑augmented pipelines.

Original Description

Jacob Lauritzen serves as the CTO at Legora, the fastest growing B2B enterprise company in history; hitting $100 million in ARR in just 18 months . Legora boasts a valuation of $5.6BN and has raised a total of $866 million in funding.
Legora's investors include the likes of Accel, Benchmark, and Bessemer Venture Partners, alongside strategic tech giants NVIDIA (NVentures) and Salesforce Ventures.
-----------------------------------------------
Timestamps:
00:00 Intro
01:19 Building Engineering in 2026 vs 2024
03:44 Code Is Now Cheap: So What's the New Bottleneck?
06:03 AI Code Review Is Still Broken
07:19 The Future Engineer: Systems Design Over Writing Code
10:28 Over 50% of Legora's Code Is Now AI Generated
12:48 How AI Is Making PMs and Postmortems Faster
15:31 Does Taste Matter or Is It Silicon Valley BS?
17:47 You Can Build Faster Than Lawyers Can Consume
31:06 What Harvey Has Done Better Than Legora
32:47 The Developer Experience Team: The Most Underrated Hire in Engineering
34:05 Hiring Engineers in Europe vs. the US
45:44 Token Maxing: Are Enterprises Using AI Wrong?
55:13 The Crazy Prediction: Lawyers Will Work One Level Above the Contract
-----------------------------------------------
Subscribe on Spotify:
Subscribe on Apple Podcasts:
Follow Harry Stebbings on X:
Follow Jacob Lauritzen on X:
Follow 20VC on Instagram:
Follow 20VC on TikTok:
Visit our Website:
Subscribe to our Newsletter:
-----------------------------------------------
#20vc #harrystebbings #legora

Comments

Want to join the conversation?

Loading comments...