We put productivity and AI software through real work, then say where it fits and where it breaks.
Real work
Real limits
Plain verdicts
Current obsession Voice input
You hold a key, you talk, and finished text lands in whatever app you were already in. We read the documentation, the pricing and the fine print, and wrote down what holds up.
Checked 28 August 2026 against the vendor's own site, pricing page and help centre. We have not run a controlled test of this product, and the page says so where it matters.
Dictation moves the work off the keyboard. It does not remove the editing.
A tool is only worth what its place in the sequence is worth. That is the difference between a review and a directory listing.
01
Getting the idea, the message or the recording into a system at all, before it evaporates.
02
Sorting and reshaping raw input into something you can actually act on later.
03
Producing the real output. The document, the reply, the piece of work someone reads.
04
Cutting the repeated manual steps, but only once the process is stable enough to be worth wiring up.
05
Catching what went wrong, and deciding what to change before the next pass.
Everything else on this list is downstream of that one question.
01
Where it sits next to what you already use, and what it replaces rather than adds.
02
Accounts, permissions and configuration standing between install and first useful minute.
03
What it costs against what it saves, including the free tier and where that tier stops.
04
Most tools suit one kind of work and get in the way of every other kind. Naming the people a tool is wrong for is more useful than another list of features, so we do that on every page.
Cut the friction
Most of the time a tool costs you is not the time it saves back. It is everything around the useful minute, and it is the part marketing pages leave out.
Seven published pages. Nothing on this grid points at something that has not been written.
Voice inputFeature
Wispr says its dictation is four times faster than typing, and publishes no method behind the number. We read what the company documents about speed, privacy defaults and compliance, and set the marketing page against its own help centre.
28 August 2026 · Feature
Voice inputEssay
Dictation has been technically good enough for years. What changed is where it sits: not somewhere you go to produce a transcript, but an input method that works in whatever you already had open.
28 August 2026 · 8 min read
Voice inputGuide
Speaking is faster than typing. That was never in dispute. The question is what the text costs you after you stop talking, and the answer depends almost entirely on what kind of writing it is.
28 August 2026 · 8 min read
MethodGuide
Six questions in a fixed order, and a two week trial built to answer them. The order does more work than the list does.
28 August 2026 · 8 min read
MethodEssay
Moving the data is the part everyone plans for. It is almost never the part that hurts, and the three costs that do are invisible until you are committed.
28 August 2026 · 7 min read
AutomationEssay
The useful line between automation and augmentation is not how clever the software is. It is whether the decision underneath the task has stopped moving.
28 August 2026 · 6 min read
WorkflowsGuide
Five stages, one tool each, and one rule about replacement. Most stacks are not too small. They are too overlapping.
28 August 2026 · 7 min read
Software reviews mostly repeat the vendor. We would rather read the documentation, find the defaults that are switched on when nobody is looking, and say plainly who a tool is wrong for.
