What is actually happening in AI, and what it changes for a business the size of yours. Written to a standing brief, checked before it ships, and sourced so you can go verify it yourself. Search it, filter it, argue with it.
Weekly digest
Every Monday, one email with whatever the archive added that week. Nothing lands in your inbox on a week the archive did not grow. One click to unsubscribe, any time.
The last 24 hours in AI, filtered hard for the things that change what a small business can do.
10 articles matching / latest 14 Aug 2026
The daily read
A twelve-person team pays about $2,880 a year for a tool whose actual job is putting a usable view on data that already lives in a spreadsheet. Google now builds that view from one sentence and writes changes back to the sheet, and it fixes none of the reasons the spreadsheet was a bad system of record.
The daily read
Every Claude model launched since August 2 weaves an invisible watermark into the text it generates, worldwide, and it travels when the text is copied and pasted. The detector that reads it has not shipped yet, which means for now the mark is a liability you carry and not a check you can run.
The daily read
The required field you configured in HubSpot has been optional for every robot pointed at your CRM: the AI agent, the nightly script, the Zap nobody owns. HubSpot said on Tuesday that changes with the September API version, and the part worth reading twice is that it changes on your schedule, not theirs.
The daily read
Meta released a 30-billion-parameter agent model on Monday under Apache 2.0 that runs the whole tool-calling loop on one desktop machine, offline. The bill it displaces is not the chat subscription. It is the loop bill, and the compliance answer that killed the project before it started.
The daily read
On August 14 Claude Code stops asking permission before every command on Pro, Max, and Team plans. Anthropic published the study behind the decision, and the finding is not that the machine got smarter. It is that in a controlled test, the humans clicking through those prompts caught a dangerous command 13.6 percent of the time.
The daily read
The free tier just got unlimited text chats and a better default model. For a forty-person shop that had been sizing a seat budget, the number that matters is now how few people actually need one, and the answer is smaller than the rep told you.
The daily read
Plugging an AI assistant into your CRM takes about four minutes now, down from a six-week engineering ticket. A scan of 25,000 of those connectors published this week found flaws in 73 percent of them, which is the first real price tag on the review that ticket used to include.
The daily read
Every content-safety tool on the market ships with somebody else's list of what counts as unsafe, and not one of those lists has "never quote a flat rate on a job that needs a site visit" on it. Mistral released a 3B model on Tuesday, free, that takes your rule as an English sentence at the moment you ask.
The daily read
The EU transparency rules went live on Sunday, and for a normal small business the whole obligation is about forty minutes of unpaid work. The part nobody is saying out loud: the moment you build the customer-facing bot yourself instead of buying one, the duty moves off your vendor's desk and onto yours.
The daily read
The flat seat you budgeted for ChatGPT Business still covers the chat box. Deep research, agent mode, image generation, and voice now draw down a credit pool the owner has to keep filled. The rate card that spells it out went live in the last day.