Reid Hoffman on tokenmaxxing: tokens measure reach, not productivity

Reid Hoffman, LinkedIn co-founder and one of the most authoritative voices in technology, has entered the discussion about so-called tokenmaxxing — a phenomenon that has been actively debated among investors, developers, and company executives in recent months. Hoffman acknowledges that tracking token consumption is useful, but warns: it is not a direct indicator of value or productivity. Tokenmaxxing is a strategy or cultural attitude toward maximizing the use of AI tools: the more tokens a team or product consumes, the more actively, by this logic, AI is being implemented.

In the business world, the idea has been adopted as a convenient way to measure how seriously a company is working with AI. Some executives already use "we burned X billion tokens this quarter" as an analog to engagement metrics in traditional SaaS. But this approach is controversial.

Hoffman has taken a balanced position. On one hand, he agrees: growing token consumption is a good signal that AI tools are actually being used, not gathering dust as a pilot project. For investors and product managers, this can be an early indicator of adoption — especially at a stage when traditional metrics have not yet accumulated.

On the other hand, Hoffman cautions against oversimplification. Tokens are not revenue, not productivity, and not a measure of work quality. A company can consume enormous volumes of tokens while generating useless content or automating processes that would be better left to humans.

Context matters: what exactly is being done with these tokens, what result is achieved at the output, and what value does it create. Critics of tokenmaxxing point to a historical analogy: the promotion of page views in the pre-internet era before the industry developed an understanding of retention, LTV, and conversions. At first, everyone counted clicks — then it turned out that clicks themselves said almost nothing.

The same thing can happen with tokens. Hoffman, as a venture investor at Greylock Partners and someone who has observed several technology cycles, is well familiar with the "vanity metrics" syndrome. That is why his warning is not merely theoretical reasoning, but pragmatic advice: use tokens as a proxy indicator for early assessment of reach, but build a comprehensive measurement system that includes business results.

For companies just beginning to implement AI, this discussion is particularly relevant. The temptation to rely on "we use AI actively — look, see how many tokens" is very strong when facing the board of directors or investors. But Hoffman's position reminds us: the simplicity of a metric is convenience, not truth.

What this means in practice: if you are implementing AI in your company — measure not only token consumption, but also what has changed. Has the time for a task been reduced, has the quality of the output product improved, which decisions are now being made faster and more cheaply? Tokens are an expense, not a result.

Confusing one with the other is a classic trap of every new technology wave.