Your AI coding tool can tell you what a session cost. It cannot tell you whether the session was worth it. Those are different questions, and the gap between them is most of what you actually want to know. The number is real, and it is thin The cost is not made up. Your tool counts tokens, and cost is derived from that count: tokens multiplied by a price. Seorak does the same, deriving the number locally on your machine from the token count, never from reading your code or prompts. So the figure is honest. A session cost what it cost. But a cost figure alone answers one narrow question, and not the one that matters. It tells you that you spent. It does not tell you whether the spending moved anything. A single number is a thin way to judge any of this, and that is not a new complaint. Developer productivity has long been treated as multidimensional, something no one metric can capture, and a lone figure is easy to mislead yourself with. Two sessions can land at the same dollar amount. One reached a clean commit and the lines it changed were still on the branch a week later. The other spun in a retry loop, never committed, and you reset the branch the next morning. The bill reads identically. The two sessions are nothing alike. A bill tells you what a session cost. It does not tell you whether the session was any good. Cost only means something next to other things A number on its own has no scale. Spending, say, half a dollar on a session is either fine or wasteful, and you cannot say which from the half-dollar. You can only say which once you put it next to what else happened in that same window of time. The same instinct holds outside this tool. The most established way to read software delivery describes it with several measures held together, lead time, change failure rate, how often you ship, how fast you recover, and it argues against judging the work by a single produced count. That is the part Seorak adds. Not a richer bill. The record around the bill. The stats it sits next to are the plain facts of the session: Duration, so you can see whether the cost came from real work or from a run that kept going long after it stopped helping. Stalls and retry loops, so a high number attached to a stuck session reads differently than the same number attached to steady progress. Whether the run reached a commit, the difference between spending toward something and spending toward nothing. Line survival, whether the lines that session changed were still on the branch later, or quietly reverted. This is the closest thing to "was it worth it," and it is the slowest to know. Put plainly, here is the same cost telling two different stories once the rest of the record is present. The values below are examples, not measured figures: Cost, Duration, Reached a commit, Lines still on branch ~$0.40, 12 min, yes, yes ~$0.40, 70 min, no, n/a Same bill. You would treat the two very differently, and you only can because the bill is not alone. The record is what you read first None of this is a verdict. Seorak is not grading the session and it is not grading you. It reads stats, not content, and it shows them sitting together so you can read them yourself. The judgment stays with you. What it gives you is a baseline. Before the longer read over time, before noticing a cost spike while a session is still running, there is just the honest record of what happened: state, duration, the token-derived cost, where it stalled, whether it committed, whether the change lasted. That record is the thing you look at first. The honest limitation is that the most useful stat here is also the latest. Line survival cannot be known in the moment. A commit happens, and only days later does the branch tell you whether those lines stayed. So the record fills in over time rather than all at once. The cost is known immediately. Whether it bought anything takes a while longer. A bill arrives the second the session ends, complete and precise and almost beside the point. The rest of the record shows up more slowly, and it is the part you were actually asking about.