The context is a new write-up and Twitter thread from the audits firm, arguing that a large class of DeFi bugs can be caught early using a simple technique borrowed from physics: dimensional analysis. The premise is blunt. If the “units” in a formula don’t match, the formula is wrong. No exploit required, no edge case needed.

That sounds abstract until you translate it into DeFi terms. Instead of meters and seconds, the system deals in tokens, prices, liquidity, and shares. These are different “dimensions.” Mix them carelessly and you get outputs that look fine on-chain but don’t represent anything real.

Trail of Bits’ point is that most smart contracts don’t enforce this discipline. Solidity doesn’t know the difference between a token balance and a price. It just sees integers. That leaves developers relying on memory and convention, which is where things break.

Reserve Protocol takes a different approach, a lesson that’s been known for millenia but often ignored to people’s peril and the destruction of space shuttles: It writes the units down.

Across its contracts, variables are annotated with both their dimension and their scale. A value isn’t just “a number.” It’s something like “unit of account per token, scaled to 27 decimals.” That might sound like overkill, but it changes how the code is read. If a developer tries to combine two values that don’t belong together, the mismatch is visible immediately.

Trail of Bits credits this as a best practice because it turns reasoning into a first line of defense, one where you don’t need to understand the entire system to spot a mistake. If a function expects token amounts and receives decimals instead, the error becomes obvious - and it is an issue that has shown up in real audits.

The firm says this as a missing habit in DeFi development, with many protocols relying on testing and audits to catch logic bugs after the fact. Dimensional thinking moves that earlier in the process and filters out entire categories of errors before they become exploitable.

It also explains why some designs work and others don’t. In automated market makers, adding two different tokens together is meaningless unless you assume they hold the same value. That’s why stable pools can get away with it, and why the same formula breaks elsewhere. The math hasn’t changed but the assumptions have.

Reserve’s approach doesn’t require new tooling or language changes but discipline - something that has been obvious to long-time followers for a while. Label what each value represents, track how it changes, and be explicit when converting between forms. It’s closer to engineering than coding.

Trail of Bits is now pushing toward automation, with internal tools that can check these relationships across a codebase. But the endorsement of Reserve is the simpler takeaway. You don’t need a new compiler to avoid basic mistakes but you do need to know what your numbers mean. It’s a structure that is doing more work than most realize.