Tomasz Tunguz, an investor at Theory Ventures, published an idea on October 23, 2025, that hasn’t left me alone since I read it: “product market fit is like United Miles. You continue to fly to maintain 1K.”

The comparison sounds simple. It isn’t. It changes how any company should think about the moment it “found” what the market wants.

Product-market fit stopped being a milestone. It’s a membership under constant review

For years, product-market fit got treated as a finish line. You hit it, you celebrate it in the board deck, and from there the work is execution: sell better, scale the sales team, optimize the funnel. That framing assumes the product that found fit stays the same relevant product six, twelve, eighteen months later.

Tunguz points at something we already watched play out brutally between 2021 and 2024: classic software companies lost their market fit “literally overnight,” in his words, without their product changing at all. What changed was what a customer could get for free, or nearly free, from a general AI tool. When ChatGPT, Claude, or Copilot raise the bar for “good enough” every few months, a product that solved a real problem in 2022 can turn, with no decision on your part, into what Tunguz calls a workflow tax by 2026.

That’s the part most product teams still haven’t absorbed. You don’t lose market fit because you did something wrong. You lose it because the market, literally, shifted underneath a product that stayed still.

Not even the labs setting the standard are exempt

What makes this more than a clever metaphor is that neither Anthropic nor OpenAI, the two companies currently defining what “good enough” means, are exempt from the same discipline. Both moved their enterprise plans from flat per-seat pricing to usage-based pricing in April 2026, and shipped newer models at higher prices within the same year: GPT-5.5 at twice the cost of GPT-5.4, Opus 4.7 at roughly 1.4 times the cost of Opus 4.6, per Simon Willison’s own accounting.

Willison ran the math on his own usage: he calculated that 30 days of direct API consumption would have cost him $1,199.79 on Claude and $980.37 on OpenAI, $2,180.16 total, against just $200 a month on the flat subscriptions he replaced. That’s not a pricing mistake. It’s evidence that even the lab with the biggest competitive edge right now is actively renegotiating how much its output is worth, in real time, instead of defending a fixed deal signed a year ago.

Tunguz uses a second example that shows the same speed from the competitive side: Google went from meaningfully behind on AI capability in early 2024 to, per prediction markets like Polymarket, the favorite to have the best model by late 2025. That full competitive reversal happened in under two years. In traditional software cycles, the same shift in standing would have taken half a decade.

What “re-earning” market fit actually means in practice

This is where Tunguz’s metaphor stops being an interesting observation and becomes an operating question. If product-market fit gets re-earned like airline miles, the right question for any product team isn’t “when did we find it?” It’s “what evidence do we have, this month, that we still have it?”

That demands a feedback loop far shorter than an annual board review. We already wrote about how the forward-deployed-engineer model solved exactly this problem in consulting: instead of handing over a recommendation and disappearing, an engineer stays embedded inside the client’s operation, watching in real time which part of the solution still generates value and which one already went stale. It’s the same discipline we documented in Uber’s agentic pods method: ten-day cycles, not quarters, to confirm what’s being built still solves the real problem.

The most common trap I see, one we already flagged in why adoption isn’t the same as transformation, is confusing the initial adoption of an AI tool with having solved the underlying problem. Adopting Claude or Copilot into your workflow isn’t re-earning market fit. It’s just the starting point for asking whether your product still delivers something those increasingly capable tools don’t yet do for your customer for free.

At IQ Source, when we evaluate whether a client’s product or service still has real market fit, we don’t review the number once a year. We treat it the way Tunguz treats his miles: the question isn’t how much you accumulated. It’s what you did this month to still deserve the status.

Measure whether your product still has the market fit you think it does