Legora CTO Calls Token‑Maxxing a ‘Really Stupid’ AI Incentive

Why It Matters

Token‑maxxing reflects a nascent stage of AI adoption where usage is equated with value. By challenging this metric, Legora’s CTO forces the CTO community to rethink incentive structures that prioritize outcomes over raw consumption. The shift could curb ballooning AI spend, especially for startups scaling rapidly, and promote a culture where engineers are judged on tangible productivity gains. If more firms adopt Lauritzen’s recommendations, we may see a wave of internal dashboards that track efficiency metrics—time saved, error reduction, cost per transaction—rather than token counts. This could accelerate the maturation of AI governance frameworks and influence budgeting decisions at the C‑suite level, reshaping how technology leaders justify AI investments to boards and investors.

Key Takeaways

• Jacob Lauritzen, Legora CTO, called token‑maxxing "a really stupid way" to encourage AI use on the 20VC podcast.
• Lauritzen advocated for reward systems based on efficiency and output, not token consumption.
• Uber capped employee AI spend at $1,500 per month per tool after overspending earlier in 2026.
• Amazon shut down an internal AI‑usage dashboard following concerns about leaderboard gaming.
• Cerebras CEO Andrew Feldman warned that unlimited token allocations are "boneheaded" and urged use of lower‑cost open‑source models.

Pulse Analysis

The backlash against token‑maxxing marks a pivotal moment in the evolution of AI governance. Early adopters treated token consumption as a proxy for innovation, but the rapid rise in model costs—especially for proprietary offerings like Claude and Codex—has exposed the flaw in that assumption. As CTOs confront tighter budgets, the metric is being replaced by efficiency‑centric KPIs that align more closely with business outcomes.

Historically, technology incentives have swung between quantity and quality. In the early cloud era, usage‑based billing encouraged teams to spin up VMs liberally; later, cost‑optimization tools forced a shift toward right‑sizing. AI is following the same trajectory. Lauritzen’s stance, amplified by peers at Uber, Amazon, and Cerebras, suggests a collective recognition that unchecked token spend erodes margins without delivering proportional value.

Looking ahead, we can expect a two‑track development: first, the proliferation of internal dashboards that surface concrete efficiency gains—such as reduced document review time for legal AI platforms like Legora. Second, a market push for more affordable, open‑source models that can be deployed at scale without the token price tag of commercial APIs. CTOs who adopt these practices early will likely secure a competitive advantage, both in cost control and in fostering a culture of purposeful AI experimentation.