A chatbot waits to be asked. An AI agent has a job and gets on with it.
The chatbot starts from scratch every time. The agent knows your business, works in your tools and does the whole job.
With a chatbot, you do the work. With an agent, you check it.
5
If you paid for ChatGPT, used it for a fortnight and then quietly stopped, that says nothing about you.
It wrote a decent email and summarised a document. On Monday the work was still on your desk.
You were sold a chatbot, and you were expecting an employee.
13
Software is now cheap and fast to build.
A system shaped around how your business works, that your staff and your AI agents can both use, is within reach of most businesses. The ones that build it first will see things in their numbers their competitors can't.
9
An AI agent can only work with what it can reach.
If everything your business knows sits in a 30-year-old system, every agent you hire starts the day locked out of the filing cabinet.
12
Visited a business turning over tens of millions yesterday. Its main system is 30 years old.
The staff have pushed it as far as it will go, and 30 years of customer and order history sits inside it where nobody can easily get at it.
6
My AI agents have started setting me tasks.
I gave one the goal: launch my book and post every day, everywhere. It came back with 7 jobs for me. 4 were logins, 2 were sign-offs and 1 was an hour a week of my time.
Exactly what a good new hire does in week one.
1
63
25bn lifetime tokens on codex, 139h longest task.
Feel like Ive been building like crazy over the last 6 months, and this is just my codex usage, Claude is more!
2
1
37
Voice agents with tools are changing how I think about getting work done. I can make progress even when I'm away from my computer—work doesn't have to wait until I'm back at my desk.
31
Who do you think will win at connect 4? Fabio (Claude Fable 5.1) Vs Grok (Build 4.7) in Workle.
I've been testing how agents running different models interact with each other, some are more verbose than others. Want to see any other models play each other?
67
In a world where software is cheap to build, every business should have its own custom software for their own unique workflows and processes…
1
3
256
I just gave GLM 5.3 a large bug discovery task, then asked Fable 5 to check its work. Every single bug verified as correct by Fable 5. GLM spend was $0.70, Fables same run $14.32. We are getting close to ridiculously low cost intelligence...
4
10
677
Ryan Edkins retweeted
I interviewed Sundar Pichai, and he made an interesting prediction for where we'll be in 3 years:
"Any content can go from any modality to any modality. It doesn't matter how the content was originally created -- people can consume it in many different forms."
"I'm very confident three years from now you will look back at what we have today and realize it was pretty primitive. It's like looking at a flip phone."
For example, YouTube clips become blogs, newsletters become podcasts without recording.
In this world, consumers would choose the format, and creators just supply the ideas, brand, and personality.
Net-net, I think this is a good thing for content.
It kills the "more content" strategy because when anyone can produce everything, volume stops being an advantage. So, the creators who win will focus on quality.
People confuse slop with AI, but they're not the same thing. Plenty of creators already use AI in their workflows and get zero hate -- people just don't realize it. Also, YouTube's new policies target slop, not AI.
If the content is good, people don't care how it was made.
I now exclusively manage my business, my software builds, and chat to my different agents in Workle. multiple projects with multiple channels, all within a single desktop software!
2
10
889
This is how you setup an AI agent to triage your email using any open source model, in just a few minutes. youtu.be/IGtBysOpKes
Loving Gemini flash 3.7 at the moment, so fast and great at fixing bugs and smaller workload tasks
2
5
291
Most builders scaling agents ask how do I add more. The better question: which one of these is allowed to argue back?
1
3
190
Two adversarial agents, eight product ideas, one instruction: kill it. Zero survived the first pass. That's the point.
2
126
A hackathon team's agent didn't write more code than the other 30 teams. It was built to attack another agent's output. It won.
2
110
Raw opposition without structure is just noise. The fix is the communication layer underneath it: defined roles, directed debate, a judge who owns the merge.
1
10





