The Math Skill in Senwitt is the one with the longest cognitive-offloading history. Calculators have been substituting for arithmetic for forty years; spreadsheets for thirty; AI for two or three. The mechanism has not changed. What has changed is the cadence — almost every cognitive act involving numbers now has a tool one click away that can perform the act for you.
This post is a deep-dive into what daily mental-math practice actually looks like in 2026, what the published evidence supports, and why the practice is worth keeping on the calendar even though no working adult needs to do long division by hand.
Why mental math is the textbook offloading case
The foundational cognitive-offloading reference (Risko & Gilbert, 2016, Trends in Cognitive Sciences) uses arithmetic as one of its canonical examples. The mechanism is straightforward — when a cognitive act has an available external substitute, the act tends to get fewer reps, and reps are what maintain the underlying skill. Arithmetic was the first cognitive act to be widely substituted by a personal tool and is the clearest empirical case in the literature.
The 2011 Science paper by Sparrow, Liu and Wegner on the "Google effect" extended the framing to information retrieval. The study found that when participants believed information would be available online later, they remembered the information itself less reliably. The mechanism is the same as for arithmetic; the cognitive act of holding the number in working memory gets offloaded onto the external tool.
The 2024 MDPI Societies study on AI and cognitive offloading (MDPI) extended the framing again to AI tools. The study found an inverse correlation between AI usage frequency and self-reported critical-thinking engagement. Arithmetic-style cognitive acts are the easiest to substitute, and the substitution pattern shows up most strongly in heavy AI users.
The deliberate-practice literature (Ericsson, Krampe & Tesch-Römer, 1993) is the constructive frame. The cognitive skill responds to daily, effortful, on-purpose engagement. For number sense, the engagement is small in volume and frequent in cadence.
What "mental math" actually means in 2026
The phrase is anachronistic if you read it as long-form arithmetic. Nobody needs to multiply six-digit numbers in their head in 2026, and the practice that defends that capability is not what the daily routine is for.
The mental-math capability that matters in 2026 is narrower and more useful. It is the ability to:
- estimate quickly — a 12% discount on £340 is roughly £40 off
- spot order-of-magnitude errors — that revenue figure cannot be right
- do a sanity check on an AI-produced number before passing it on
- carry a few key figures in working memory long enough to compare them
- feel whether a quoted statistic is plausible
This is number sense, not arithmetic. The cognitive act is judgment about quantities. The substrate that supports the judgment is daily contact with calculation. The substrate weakens when the daily contact disappears.
A daily mental-math practice that fits a working life
The routine below takes about five minutes a day and does not require special equipment. None of it requires you to stop using a calculator or AI for the calculations that genuinely benefit from them.
1. Estimate before you calculate. When you have a calculation to do, do a rough mental version first. Then do the precise one. The estimation is the rep; the precise calculation is the check.
2. Spot the order of magnitude on every number you see. Quoted statistics, AI outputs, news headlines, spreadsheets. Is this number in the range you would expect? The act of asking the question is the rep, even when the answer is obvious.
3. Do one mental calculation a day. A real one, not a trivial one. The change from a £20 note on a £6.40 bill. A 15% tip on £67. The percentage of a number you actually need to know. Do it in your head before reaching for the calculator. The fact that the calculator is available is the point of the discipline.
4. Sanity-check AI numbers. When AI gives you a calculation result, hold it long enough to feel whether it is plausible. The verification step is the rep that protects against the Mata v Avianca equivalent in numbers — accepting an AI-produced figure that is wrong by an order of magnitude because nobody bothered to check.
5. Carry one figure across a conversation. When you read a number that matters, try to hold it in working memory through the next few minutes of the conversation. Working memory is itself a cognitive act, and it weakens when every number gets immediately offloaded to a note.
What this is not
A few honest disclaimers.
This routine does not claim that calculator use or AI use causes general cognitive decline. The published evidence does not support that claim. The narrower claim — that cognitive acts which get fewer reps tend to weaken, and that arithmetic and number-sense acts get fewer reps when tools substitute for them — is what the cognitive-offloading literature supports.
It is not a case for refusing calculators or AI for calculation. Both are useful and the productivity case is real. The recommendation is calibration: do the cognitive act yourself when it is small enough to be worth doing, use the tool when it is not.
It does not promise that daily mental-math practice transfers to general intelligence, mathematical ability beyond the practised range, or any other cognitive surface. The deliberate-practice literature is consistent on the specificity of transfer — the practised skill is what improves, not unrelated skills. Number sense is what the practice maintains. That is the scope.
How Senwitt fits
Senwitt's daily Set includes a Math Skill rep most days. The reps are short — mental arithmetic, estimation, sanity-checking — and are designed to maintain the number-sense substrate rather than teach mathematics. The deliberate-practice frame is taken from the Ericsson 1993 paper.
The research/cognitive-offloading page covers the foundational research. The research/google-effect-and-digital-amnesia page covers the related literature on offloading and recall.
A practical note on what working professionals tell us about the routine after a few months of doing it. The single biggest effect is not increased speed at arithmetic, which is not the point. It is the regained ability to feel when a number is wrong before the AI or spreadsheet has been re-checked. The story we hear most often runs something like this: an AI assistant returns a figure during a meeting, the practitioner pauses for two seconds, says "that does not feel right, can we re-run that," and is correct often enough that the habit pays for itself in avoided errors several times a month. The cognitive act being protected is not arithmetic. It is the order-of-magnitude alarm — the quiet voice that goes off when something is wrong by a factor of ten. In a working environment where most of the numbers passing through your day are AI-mediated, that alarm is the part of the role that AI cannot do for you. Daily practice is what keeps the alarm working.
