The brain-exercise category has a credibility problem, and it earned it. Through the 2010s, several large companies in the space made broad cognitive-improvement claims that the published evidence did not support. The result was regulatory action, a high-profile consensus statement from the scientific community, and a generation of skepticism that hangs over the category to this day.
Senwitt is built inside that context. We are deliberate about what we will and will not claim, because the gap between the marketing instinct and the evidence is real and it matters. This post walks through what the literature actually supports, where it stops, and how that maps onto Senwitt's claim line.
The narrow claim the evidence supports
Two literatures matter here.
The first is the cognitive-training and brain-training literature. The most-cited synthesis is Daniel Simons and colleagues' 2016 review in Psychological Science in the Public Interest — roughly 200 pages and the closest thing the field has to a consensus document. The careful summary of what brain-training programs can be shown to do:
- Practising a specific cognitive task produces improvement at that specific task. This is the most reliable finding in the entire literature.
- There is modest transfer to closely related tasks — tasks that draw on similar underlying processes.
- There is limited and inconsistent transfer to general cognitive outcomes — broad measures of intelligence, real-world cognitive performance, prevention of age-related decline.
This is not a dismissal. Practising specific thinking skills genuinely improves those skills, and that improvement is worth something on its own. The dismissal is of the bigger claim that practising puzzles in an app translates to broad cognitive enhancement or decline prevention.
The second is the cognitive-offloading and skill-disuse literature — newer, sparser, and more directly relevant to the AI moment. Risko and Gilbert's 2016 Trends in Cognitive Sciences review, the 2011 Sparrow et al. Google-effect paper in Science, and the more recent work on AI-mediated thinking (including the MIT Media Lab cognitive-debt 2025 preprint on arXiv, with the Stanković 2026 critique also in preprint) point toward a complementary finding. When cognitive work is delegated to external systems — search engines, GPS, AI assistants — the encoding pattern shifts and the underlying skill gets fewer reps. The 2026 Anthropic study on coding skill formation found a measurable difference in skill mastery between AI-assisted and unaided conditions.
Put the two literatures together, and the narrow defensible claim is roughly this: regular practice of a thinking skill maintains that skill better than non-practice; in an environment where AI and other tools are doing more of the everyday cognitive work, the daily existence of unaided practice is more relevant than it used to be.
What the evidence does not support
The same literatures are clear about what they do not establish.
They do not support broad cognitive-decline-prevention claims. This is the claim line that the 2014 Stanford Consensus Statement and the 2016 FTC Lumosity settlement were specifically about. The evidence did not support "our app prevents age-related cognitive decline," and the regulatory action was a direct consequence.
They do not support IQ-improvement claims. The intelligence literature is large, careful, and not friendly to the claim that any app or short daily practice can move general intelligence in a meaningful way.
They do not support clinical claims. No app should claim to prevent or treat dementia, Alzheimer's disease, or any other clinical condition unless it has been through the relevant regulatory process. None of the products in the brain-exercise category have. Senwitt has not, and we do not make those claims.
They do not support "makes you smarter" or "boosts your brain" framings. Both phrases are loose enough to imply far more than the evidence supports, and both are now closely associated with the marketing patterns that drew regulatory action. We avoid them in our copy.
What Senwitt actually claims
Senwitt's position is narrow on purpose. We claim:
- A daily, mixed Set of short, deliberate, unaided cognitive practice across writing, math, code, memory, reading, and reasoning gives those specific skills regular reps.
- In an environment where AI tools are doing more of the everyday cognitive work, the existence of unaided practice somewhere in the day is increasingly relevant — and that's what we're built for.
- We track a self-reported Sharpness signal so that users have a private read on their own pattern, not so that we can claim population-level cognitive outcomes.
We do not claim our product prevents cognitive decline. We do not claim it improves general intelligence. We do not claim it protects against dementia or any clinical outcome. We do not make brain-improvement promises beyond the deliberate-practice base rate that the literature supports.
The reason for the discipline is straightforward. The category has a history of overreach. The cognitive-offloading argument we do make is strong enough to stand on its own. Adding the broader claims would undermine the narrower, defensible ones — and would put us on the wrong side of the same evidence line the Stanford Consensus and the FTC drew in the 2010s.
Why the AI angle matters newly
The Senwitt position is not a brain-training position re-packaged. The argument is structurally different.
The brain-training claim was: "our specific puzzles improve your general cognitive performance." That claim did not survive the evidence.
The cognitive-offloading claim is different. It is: "when more of your everyday thinking is mediated by external systems, the underlying skills receive fewer reps, and the evidence for skill-specific atrophy is real." That claim is in the literature, and it has been growing as AI use has spread. The right response to it is not a transfer-of-training story; it is a deliberate-practice story applied to thinking skills the same way it has long been applied to musical, athletic, and craft skills.
In other words: brain training was selling cognitive enhancement. Senwitt is selling deliberate practice for skills that the AI environment is putting in less daily use. The framing is narrower, the evidence is more direct, and the claim line is honest.
Why the language matters
The phrasing we use across the site reflects the claim line above. We say "keep your thinking skills sharp," not "make you smarter." We say "deliberate practice," not "brain training." We say "daily reps," not "cognitive enhancement." When we cite preprints, we flag them as preprints. When we summarise a study, we describe what it actually measured rather than the headline version of what the press release implied.
This is partly editorial discipline and partly a structural feature of the position. The deliberate-practice case is strong enough on its own evidence. The cognitive-offloading case is real and growing. The combination is enough to build a serious product around without overreaching. Overreaching would put us in the company of the 2010s brain-training industry — and that is not a place we want our credibility to live.
The research/scope-of-evidence page is the formal statement of all of this. It is one of the most important pages on the site, and it is the one we update most carefully when new research lands.
