Four small tools, assessed rather than promoted

What the Pomodoro technique is actually doing

Francesco Cirillo devised it as a student in the late 1980s, timing himself with a tomato-shaped kitchen timer. Twenty-five minutes of work, a five-minute break, and a longer break after four rounds.

The evidence for scheduled breaks improving sustained performance is reasonable. The evidence that twenty-five minutes is the right number is nonexistent, because it was never derived from anything — it was the interval on the timer he had.

• It lowers the cost of starting. Committing to 25 minutes is much easier than committing to a task, and starting is where most procrastination lives.
• It forces breaks that would otherwise be skipped.
• It makes effort visible, which helps with estimating.
• It gives interruptions a defence — "I will come back to you in twenty minutes" is a complete answer.

Those are real benefits and none of them depends on the specific number.

Where it does not fit

The honest limitation is that a timer interrupts, and some work is damaged by interruption.

Deep problem-solving has a long ramp-up, and being pulled out of it at a fixed interval is a cost rather than a rest. Research on task switching consistently finds substantial time lost to regaining context after an interruption — the widely cited figure is over twenty minutes, and while the exact number varies by study and task, the direction is not in dispute.

So: use it for work you resist starting, for shallow or administrative tasks, and for anything where you lose track of time. Set it aside when you are in flow — finishing the thought and taking a break afterwards is better than obeying the timer. And adjust the interval. Fifty minutes with a ten-minute break suits deeper work; the method is not diminished by changing the number.

Why people cannot pick at random

Asked to produce a random sequence, people avoid repetition. Genuine randomness contains runs — a fair coin produces four heads in a row about one time in sixteen — and human-generated sequences contain far fewer than they should.

The same bias affects picking a name from a list, choosing a number, or deciding who goes first. People also favour the middle of a range and avoid the ends.

• Drawing from a hat is decent if the papers are the same size and genuinely mixed.
• Physical wheels and dice can be biased by wear and manufacture.
• Sorting by a random key is a classic error — it produces a non-uniform shuffle. Use Fisher-Yates.
• Modulo of a random integer is subtly biased unless the range divides evenly.
• Math.random() is not cryptographic. Fine for a prize draw, unsuitable for anything security-related.

For anything consequential — a prize draw with real value, an audit sample — the process needs to be defensible as well as fair, which means documenting the method and ideally the seed. Spin the Wheel covers the ordinary cases where the point is a visible, unarguable choice rather than cryptographic rigour.

Recording the screen with nothing installed

Every current browser can capture the screen through a standard API, and the operating systems have their own recorders too. Installing software for this is rarely necessary.

The browser route has a useful property: because the capture happens in the page, a recorder built this way can keep the recording entirely local rather than uploading it.

Two practical notes. Choosing a single tab rather than the whole screen avoids accidentally recording notifications, and it is the option worth defaulting to. And system audio capture is separate from microphone capture, is not offered on every platform, and is the reason a recording turns out silent — check both before recording anything long. Screen Recorder runs in the browser and keeps the file on your machine.

Proofreading by ear

This is the most effective proofreading technique that costs nothing, and it works because of how reading fails.

Reading your own writing, you see what you intended. The brain supplies missing words, corrects transpositions and skips duplicated ones — familiarity is precisely what makes self-proofreading unreliable. Hearing the text removes the visual shortcut.

• Missing small words — "the", "of", "to" — which the eye inserts and the ear does not.
• Duplicated words across a line break, almost invisible when reading.
• Sentences that run too long, obvious when a voice has to get through them.
• Awkward rhythm and repetition of the same construction.
• Wrong-but-real words, where the sound is close enough to catch some cases.

Slow the speech slightly below a natural pace and follow along in the text. macOS has Speak Selection, Word has Read Aloud, and browsers can do it natively — Text to Speech handles pasted text without installing anything. It is the one proofreading step that reliably catches errors a spell checker cannot see.

Frequently asked questions