Introducing Workotron 1.0.
Your private 24/7 AI computer in the cloud, with a chief of ops that runs your business for you.
70 workflows. Drives all your web apps. Smarter every day.
In private beta. Message me if you'd like to try it out.
https://t.co/O4AHpN6KZx
Here's how we designed this:
1. We identified the most useful and common small business use cases based on web services popularity and authored the workflows for them.
2. We built an execution engine that runs workflows reliably with all the context they need, execution includes output verification before results are surfaced to users.
3. We implements a large number of workflow evals to gain confidence that workflows work well and shipped them.
4. We have built a self-improving workflow that improves workflows after each time they run automatically.
5. We've built an update mechanism that smartly merges changes to workflows from template updates.
We are today at 78 templates, but Workotron can create workflows for users for use cases that are not covered by our templates based on patterns in the existing templates.
Introducing Workotron 1.0.
Your private 24/7 AI computer in the cloud, with a chief of ops that runs your business for you.
70 workflows. Drives all your web apps. Smarter every day.
In private beta. Message me if you'd like to try it out.
https://t.co/O4AHpN6KZx
@kachersoft I haven't noticed any performance degradation after fast compacting and I've been working with it for many months. Try it out for few days. You can always go back to /compact
Every Claude Code compaction cost me a minute and a half, several times a day. So I built fast-compact. It does the same job in half the time, and I've used it for months. How it works, the numbers, and the gist 👇 https://t.co/7KdhN9iu8h
@AppsolutelyxD If you remember to /clear after /fast-compact then you don't need that at all. This is for rare cases that you forgot to /clear - you can make it yours and adjust as you see fit :-)
I spend 10 to 16 hours a day with Claude Code and Codex. By noon the walls of text wore me out, so I built action cards. Here's how to build your own 👇 https://t.co/aRRvkDOpcB
I've been working with fast compact for many months and I never ran into an issue of the session lacking context or having any degradation of performance once I got it to the final form that I've shared here today. HAve a look at the gist for what it keeps. You can easily modify the skill to remember additional things.
Have a look at the skill report template in the gist. What's cool is that you can customize it to your own needs easily by just editing the skill. I found that writing great skills is an iterative process - what you see in the skill is the results after many versions until I got it right for me - meaning, I never notice any degradation perf post-compact in my own sessions.
This is cool but overly complex and the next level is chief of ops. You only need chief of staff when you have staff to manage. Chief of ops manages tasks. Jobs and work are made of them and there's no need for an AI workforce when your chief of ops can execute, monitor and delegate to agentic tasks managers. We don't need an AI workforce, we need an AI chief that can get work done.
Cool but these days this is a bit of a legacy way to look at QA. QA by agents should run 24x7 before, during and after a launch. The old paradigms are outdated. They should also delegate obvious low-blast radius bugs to coding agents who fix over night so tomorrow's release is already better. Non-obvious bugs should be surfaced for the day's sessions.
@aakashgupta@thejamesmcaulay No multiple tabs needed anymore. You can do all of this only in one tab and keep full control while automating many hours of working across different apps and tabs via Workotron. You may want to have a look at it.