The newsletterOnce or twice a week. 1 hot take, 2 tool suggestions, 3 best practices.
No. 006
Pretty soon you won't know what software you're using.
I’ve had treg connected to Claude for about six weeks. I wrote about it last issue. It’s one account that lets Claude call a few thousand paid tools, and I pay a few cents each time it uses one.
I was telling someone about it this week and realized I can’t name the tools I’ve used through it. It’s probably 10 to 20 of them and I’d have to log in to my account to check. Claude gets a list of tools and uses whichever one makes sense in the moment.
I’ve never used software before that I didn’t know I was using.
OpenRouter did this for AI models first. It’s one account that gets you to pretty much every model, and you pay for what you use. treg self-declares as the OpenRouter for tools. And I think a lot of software is headed the same way. You won’t pick it. Your agent will pick it from a list, and you’ll pay based on usage.
Think about what that does if you sell software. The buyer never sees your interface, sure, but also never sees your website, branding, or even knows your name. If treg adds 10 new tools tomorrow, Claude just starts using them, and one day I wake up and my AI is more powerful and I don’t know why. And if two tools do the same job, the agent is going to pick the cheapest one.
I’ve been asking the early stage companies I work with the same question. Why do you have an interface? You could put the same functionality behind an MCP server (the connection an AI tool uses to call your product) and deliver it a lot faster.
Curating tools to power up agents and save people on subscriptions could be a huge category we start to see explode.
The software game is going to change if this becomes the way people are interacting with software, because their agent’s choosing the software for them on a per use basis.
Two tools I’ve been trying. One kind of tool I didn’t end up needing.
Muse
Muse is Meta’s new personal agent. It gets its own computer in the cloud with its own browser, and you can watch it work.
It failed at the first thing I gave it to do, but it wasn’t its fault. I switched to AT&T nine or ten months ago and T-Mobile has been charging me for a line I don’t have access to the whole time. I gave Muse my login and it got into an argument on T-Mobile’s support chat for me. It didn’t get anywhere. But it had no problem navigating to where it needed to go to start talking to someone.
The browser use stuff has gotten ridiculous. If somebody on your team is clicking buttons in a browser because a system doesn’t have an API, I can’t see a reason to pay a person for that anymore.
YouSpot
YouSpot is a new one from HubSpot, built for one-person companies. It builds what they call a second brain out of your email, your LinkedIn and whatever else you connect, and then Claude or ChatGPT can pull from it.
It’s been interesting to use, and for the price it’s an interesting take on a second brain. It’s pay as you go, the same model as the hot take, and I don’t think I’ve put more than $1 into it.
It’s not going to replace Notion for me. But I’m curious where it will go.
Clip editors (OpusClip, Descript)
I may change my mind on this one. I was breaking out clips from my recent podcast with Scott Ward and expected to use something like OpusClip or Descript for it.
I didn’t need to. Claude had no issue making the clips and adding the captions for me, and I got great results from it editing the video directly.
Replacing it with
Three things I'd do this week
Check which model your team is using to write emails.
I’m doing AI enablement for a client and one of the first things I noticed was that a third of their usage was on the most expensive model. Turns out people were using it to write emails. Why is that a problem? Imagine you have a loose screw. Are you going to twist it in with your finger, grab a screwdriver, or roll in the industrial robot? Pull the usage reports and dive into your users making heavy use of the more expensive models.
A lot of it does come down to what you’re connecting it to in the context and data it has access to more so than the model itself.
Make the company AI account better than the personal one.
Everybody has their whole life of context built up in their own personal account. You can’t just mandate the company version. They’re not going to use it. You have to make the company account the more useful one, with the shared context, the connections and the instructions already in it. That’s the thing I’d be focused on right now.
You have to make it more attractive to use the company version of the thing than the personal version of the thing. You can’t just mandate it.
Fix the sales to onboarding handoff first.
It’s the handoff that gets the least attention and it’s always a disaster, because no one is thinking about what the onboarding team needs to do their job. Then somebody patches it with a meeting. Now the rep has to sit down with the onboarding person before every kickoff, which is the least efficient way to move information I can think of. Ask onboarding what they need, then collect it during the sales process.
That’s always a disaster because no one’s thinking about what the onboarding team needs to do their job effectively.
Thanks for reading. If your AI is already using tools you couldn’t name, reply and tell me how you found out.
Kevin
The newsletter
This one went out in October 2026. Get the next one first.
Once or twice a week. Quotes from the calls, the tools I would buy, and what I would do this week.
No spam. Unsubscribe in one click.
Generate revenue. Subtract everything else.
Leads slip through cracks, reporting is a mess, and growth feels stuck. You’re not missing tools. You’re missing the system that ties them together.