AI tools are now sharing each other's plugins. Here's what that actually means for you.
This one is a plumbing story, not a business story, and the short answer is: you do not need to do anything. A developer has built a bridge that lets one AI coding tool borrow another tool's library of add-ons — commands, skills, connections to other software — without switching between them. Small thing. It makes headlines in tech circles because it hints at where these tools are heading. For a plumber, a salon or a two-person agency, it changes nothing this week. But the reason it exists is worth five minutes, because it tells you what to expect from any AI team you pay for, including this one.
What actually happened
A developer released a small connector that lets a second AI coding assistant install add-ons built for a different, more established one — pulling straight from that first tool's official library. It is not a new product. It is a translation layer, built by one person, that lets tool B use tool A's collection of extras. The author was upfront that not everything survives the translation: some pieces only partly work, some get dropped entirely with a warning. That candour is the most useful part of the story. It is a young bridge, built fast, with known gaps.
None of that is a decision you need to make. You are not choosing between coding tools. You do not run a coding tool at all. This is one piece of software talking to another piece of software, several layers away from anything on your desk.
Why it made the news anyway
The interesting part sits underneath the connector story. Several people this week talked about turning a book, a playbook or a set way of working into something an AI assistant remembers automatically — so a person does not have to re-explain the same framework every single session. One well-known product leader described building herself a set of these from management books she uses often. A separate toolmaker shipped an update so those reminders install once and then update themselves, rather than needing to be reinstalled by hand each time.
Strip away the developer language and the pattern is simple: people are tired of teaching the same thing to their AI tools over and over, and they are building ways to make a lesson stick. That is a real problem. It is just not one you have, because it is already solved in the product you are paying for.
The bottleneck this actually touches
If you run a small business, the real version of this story is not "which AI tool should I install." It is: work only happens when you are the one sitting at the desk explaining it. Every time you bring on a new tool, a new template, a new way of writing an email, you end up re-teaching it — to the tool, to a new hire, to yourself six months later when you have forgotten the exact wording you wanted. That re-teaching is the tax. This week's news is developers building ways to pay that tax once instead of every time. It is worth knowing the pattern exists, because it tells you what to expect from any AI product that claims to work for you rather than just respond to you.
WorkMate was built around exactly that expectation from day one, which is why nothing in this week's news changes what your crew does. Your brand kit — your voice, your tone, the words you don't use — is stored once and applied by every mate on every task. A correction to how invoices get chased, or how enquiries get answered, sticks between sessions without you saying it twice. That is not a feature bolted on in response to a trend. It is the reason the crew feels like it is actually learning your business instead of starting fresh each morning.
A caution worth carrying
One thread running through this week's coverage is worth repeating plainly: several people who build these AI add-ons for a living admitted, in public, that nobody has actually measured whether they change the AI's behaviour for the better. They are unproven by the people selling them. That is a healthy scepticism to hold generally — if someone offers to sell you a custom "AI upgrade" for your business, ask what changed and how they know, not just whether it installed. It is also why draft-review-send stays the shape of outward-facing work here: your crew produces the draft, you approve it before it goes out, and nothing gets sent on a promise alone.
What to actually do this week
Nothing about your setup needs to change. If you want the underlying idea to work for you, it already does — the next time you correct how a mate writes a follow-up email or handles a booking query, that correction is remembered without you filing it anywhere. That is the whole point of a shared Brain: the lesson only has to be taught once.
Related reading: One Brain, never off the clock and Not another chatbot.