For emails that keep coming back, like declines, rescheduling and information requests, let AI draft the reply, but a person must first decide the promises: deadlines, exceptions and what is possible. We gave three made-up photo studio emails to AI six times in two ways. Every calculation was right, but four of the six drafts contained a promise or judgment the owner never made. We share three reply templates and three checks before sending.
Read now →Jensen Huang, who led NVIDIA for 33 years, stepped down on November 11, 2026. We sort out why he left, how far the spy allegations have been confirmed, who is running the company, and what it means for the stock and for Korean chipmakers, splitting everything into filings, announcements and reports, and adding our own interpretation.
On November 12, Google released Gemini 4 Hydrogen, a fast everyday model. Using only Google's official announcements, we cover what changed compared with the previous Gemini 3.8 Flash, how Google Search changes, where free users, paid users and developers can each use it, and what it costs.
We follow a day in the life of a one-person business owner with no staff and look at five tasks worth handing to AI (quote replies, expense notes, notice rewriting, to-do lists, idea sorting), plus where a person should check each one. We actually tried three tasks for a made-up workshop owner, including one where the same '10 business days' gave different due dates depending on which day you start counting from.
On 2026-11-13, using paid subscription accounts, we gave the command-line programs of three AI companies the same three tasks: turn meeting notes into a table, work out an order total, and quote a company rule that does not exist. We did not check the free plans or the web apps this time. All three got the math right. With these programs and settings, the differences were in whether they kept the notes' wording and what they added when they did not know. This post covers the actual answers, our own math check, four questions to ask when choosing, and a test sheet you can run yourself.
When AI tidies a spreadsheet, rows can vanish or numbers can change and the table still looks clean. This post covers what to check before you paste a table in, a prompt that asks for one fix at a time, the three checks (row count, total, 5 sample rows), how to find changed cells, and a record of two cleanups of a made-up order sheet compared against the original.
When you give meeting notes to an AI, the sentences read smoothly, but decisions, to-dos, owners, and deadlines often go missing or get invented. This post covers three steps: paste the notes as they are, set the columns and get a table, and decide where you will check for gaps. It also has a prompt you can copy and the results of running a fictional meeting twice.
Pick tasks for AI not because they look like something it would do well, but because a mistake would be easy to spot. Answer five questions, add up the score, and each task lands in one of three groups: hand it over now, hand it over with checks, or don't hand it over. We scored 10 fictional office tasks and include the results.
When an AI gives a confident answer that might be wrong, here are five ways to check it, in order. It includes a table for choosing how many to use based on how risky the question is, and a record of checks we ran ourselves against public sources.