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.
31 articles matching / latest 30 Sept 2026
AI News
OpenAI spent Tuesday announcing more than twenty things. Two of them change what your subscription is: sixteen partner products can now draw on your ChatGPT allowance, and an always-on agent draws on it too. The same plan holds half as much from October 30.
AI News
In four days the cost of having an AI do a unit of work fell by roughly half at two labs, Amazon opened its seller platform to outside agents, and Salesforce said the AI is replacing its interface. The same four days produced a government portal an agent let itself into and an on-the-record admission from the people selling all of it that nobody is steering.
Git Articles
A drafted floor plan runs $500 to $2,000, and the do-it-yourself route is a SketchUp Pro seat at $399 a year. Pascal is an MIT-licensed 3D building editor that opens in a browser tab with one command and takes instructions from the coding agent you already pay for. It also has three contributors and a stable release from June.
AI News
Meta shipped an agent that stores your logins where the agent itself cannot read them, asks before it spends, and issues a single-use card locked to one merchant and one amount. The control set is the actual news, and it is now the checklist you run against every agent vendor that calls you this quarter.
AI News
Anthropic shipped a way to keep an AI agent's instructions in a text file, with an approval step before a change takes effect and a diff after it does. The thing it replaces is the automation nobody can explain and nobody dares turn off.
AI News
HubSpot now ranks every open invoice by how much of your money is at risk and writes the chase email underneath it, in every portal, on every tier. Ranking was never the part that was hard. Nobody owning the follow-up was.
AI News
GPT-6 Astra scored 97.6 percent on frontier mathematics and 41.4 percent on AutomationBench, the benchmark built out of two billion real Zapier tasks a month. In April the best models were under ten percent. The distance between those two numbers is the whole planning problem.
AI News
Anthropic cut the price of re-reading context on Fable 5.1 by 75 percent, down to a fortieth of what a fresh read costs. The saving lands in proportion to how tightly your work is packed in time, and the penalty for breaking a cache quietly quadrupled on the same day.
Foundational AI: Do's and Don'ts
A free trial hands you the software and none of your data, and your data is the only part that was ever in question. Here is the two-hour test to run before you open the trial, and the one number that decides the purchase.
Monthly State of GTM
In one month the human approval step was deleted from the tool you build with, from the automation that emails your customers, and from the market where you used to buy review by the item. Something went in where each one had been, and it was a meter.
AI News
You can buy a support agent today that charges 99 cents when it finishes a conversation and nothing when it hands off to a person. On Monday OpenAI started testing the same deal with its biggest customers. The catch is that somebody has to define finished, and it is not you.
AI News
Amazon's own pricing example had 200,000 scanned form pages read by three different people each, every month, for $12,200 and no hiring. On September 30 that stops being something you can buy, and the reason is not that it stopped working.
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.