Justin Roff-Marsh

Buffett really aged in the past year. His retirement wasn't random

GET RICH QUICKGET RICH QUICKGET RICH QUICKGET RICH QUICKGET RICH QUICKGET RICH QUICKGET RICH QUICKGET RICH QUICKGET RICH QUICKGET RICH QUICKGET RICH QUICKGET RICH QUICK

Eli Goldratt's book, The Goal, was famous for its (then unpopular argument) that keeping every machine running 24 hours a day, the metric most plant managers cared about, was actively making factories worse. I suspect we're seeing the same fallacy in how many people are using AI agents. Goldratt's point was that machine utilization isn't throughput. What you want from a manufacturing plants is making good widgets as cost-effectively as possible. It doesn't necessarily follow that running your machines all the times optimizes that. Picture a three-station assembly line. Stations 1 and 2 each crank out 200 widgets an hour. Station 3 can only handle 100. Running stations 1 and 2 around the clock doesn't ship more product. It just piles up half-finished widgets in front of station 3, ties up cash in inventory, and creates more work managing the pile. He developed the Theory of Constraints to point out that what matters is solving the bottleneck in the system, not increasing machine utilization. I suspect a lot of agent usage right now is the same fallacy at higher resolution. Running 20 Claude Code sessions in parallel can feel productive because something is always happening. But, if the bottleneck in your work is judgment about what's worth doing, more agents just generate more output for you to wade through. This is not to say there aren't workflows running 20 agents in parallel very effectively, I'm sure there are. And, I suspect there's a general retraining we all need to do around evolving historical workflows. But.... The constraint for most knowledge work is deciding what's worth executing and no one is task switching between 20 things at the same time effectively I don't think. I find I can run maybe 2 or 3 things in parallel with maybe 1 or 2 admin-y type things on the side and that is only if I'm very locked in.

Musings on Throughput per Constraint Unit. In the TOC community, our most critical ratio is Throughput per Constraint Unit (T/Cu). I’m not convinced everyone understands the power of this metric. If you’re considering the past, profit is arithmetic (revenue minus cost). But if you’re predicting the future, profit is calculus (the difference between the rate at which Throughput is generated and the rate at which operating costs accrue). This means that if we want to make management decisions intelligently, calculus is required (and calculus is difficult). But here’s the thing. EVERY business, when forced to operate in an orderly (i.e., non-chaotic) state, assumes exactly the same configuration of resources. You have one Constraint (fleet of aircraft, in the case of an airline) that is consistently heavily loaded, and a large number of non-Constraint resources, all of which have protective capacity (baggage handling, crew scheduling, etc), enabling them to subordinate to the Constraint. This configuration means that there is a SINGLE resource that determines the rate at which the organization generates profit. This simplifies predictions and, therefore, management decision-making. Given that profit is the difference between the rate at which Throughput is generated and Operating Costs accrued—and given that the rate of the former is determined by the Constraint and the latter is reasonably constant—predictions can now be made with simple division. delta(T/Cu) ∝ delta(Profit) In other words, changes in T/Cu are proportional to changes in profitability. Throughput is pure contribution margin (revenue minus raw-material costs). But where are Operating Costs in this equation? A Constraint Unit is a unit of the organization’s capacity (remember, the Constraint is the pacesetter for the entire organization). And the organization’s capacity is exactly what your operating costs are paying for! Assuming that operating costs are fixed for the period under consideration, T/Cu enables managers to make surprisingly accurate predictions with minimal effort. It’s just as simple as the traditional cost-accounting approach to predictions—and it avoids the nonsense results.