The reading room · Digna Legi
TBM 437: Tokens, Hours, Points, and Other Curious Proxies
/100
80–100: high value. 70–79: worth the time. Below 70: below the usual publication threshold.
Evidence-reviewed score based on available publisher text. Evidence is sampled with gaps, so omitted sections may affect balance or completeness.
Scores reflect one reader’s profile, not an objective quality rating. Best is a separate personal selection.
How scoring works →This brief · about 3 min with detail
Why read this
Precise input meters can make ROI look rigorous while hiding the untested theory of value that actually determines return.
AI brief · Checked against source text
The main idea
The central warning is that tokens, like hours, capacity, and story points before them, are attractive because they are easy to count, not because they explain value. The author argues that ROI only becomes defensible when the cost unit is tied to a causal model: what demand is served, what capability is created, what bottlenecks move, and what assumptions would change the investment.
Go a little deeper
Inputs are not interchangeable capacity
Hours look concrete but collapse crucial differences: timing, skill context, system constraints, setup costs, and long-term option value. The same nominal hour can unlock a blocked system, disappear into context switching, or quietly improve future speed by reducing codebase debt. Treating time as a fungible allocation unit makes the investment side look precise while erasing the production mechanism that creates or destroys leverage.
Flow metrics need stable work classes
The author does not reject lean or Kanban-style measures; he limits their job. Cycle time, throughput, work-in-progress limits, review latency, and merge frequency can diagnose bottlenecks and waste when work types are segmented. They become misleading when heterogeneous product discovery, experiments, maintenance, and system-changing work are averaged into one stream, because changes in the metric may reflect mix, not performance.
Team-level funding shifts rigor to causality
Instead of decomposing a team into hours, points, or feature-level cost attributions, the proposed alternative treats the team as the investment unit. The harder work then moves to maintaining a business case: the demand the team serves, the proxies believed to predict differentiated value, the uncertainties, and the evidence that would increase, decrease, or redirect funding.
Tokens intensify the measurement illusion
AI spend adds a new twist because token consumption produces an unusually exact meter inside an uncertain economic system. A cheap generation step may create downstream review, maintenance, coordination, or cognitive load; it may augment people, replace tasks, or merely move the bottleneck. The serious question is therefore not token ROI in isolation, but what human-plus-AI production system the organization is betting on.
A case from the article
The $50 million AI spend comparison
The author imagines finance approving $50 million of token-based AI spending and compares it with the same investment in human capital. That comparison exposes what token accounting misses: employees create institutional memory and relationships, AI consumption may or may not leave residual assets, early use cases may outperform later marginal ones, and both human and AI production can overload downstream systems.
How the case is made
The case is made through measurement critique, manufacturing analogies, and a practical taxonomy of AI-spend assumptions.
Where the idea has limits
The argument applies most directly to product, software, and AI investment systems where work is exploratory, interdependent, and hard to attribute to individual units.
A question to take away · from Digna Legi
Where are you measuring a clean denominator while leaving the value theory vague?
What the original adds
The source includes a fuller comparison of hours, flow metrics, story points, team funding, and token spending, plus concrete hypothesis templates for AI business cases.
About this brief
AI-written, then separately checked for source support, useful detail and clarity. The author’s claims and our editorial question are kept separate. The original remains the author’s work. How we select and summarise →
How was this brief?
Rate this summary, separately from the author’s article.
Optional. Saved in this browser; shared only if you allow analytics.
How was the original article?
Rate the author’s original after reading it.
Optional. Saved in this browser; shared only if you allow analytics.
Digna legi. Worth reading.