Social media is a 3 headed dragon. It encourages a lack of focus, which encourages those negative thoughts to keep playing in your head, which then encourages you to believe them because you lack focus to take action. The only way to kill negative thoughts and improve mental health is to act. Social media doesn’t want you to act.

Heard this recently: The devil knows your name, but calls you by your mistakes. God knows your mistakes, but calls you by your name. When someone comes into your life and talks about your mistakes, you know where they are coming from.

Vibe Coders! I can help level you up so what you create looks like it was created by a team of senior software engineers. I think people are calling this graph engineering, but i really dislike that term. I've been running agentic engineering teams for a while that deliver senior level results and in the last few months, i have been extending this out. I like to think of this as my personal command center where I can pull in any team (engineering, marketing, copywriting, etc) to work on something. In the image below, this is all working within the claude code environment today. BUT, I am building the conductor as a proper desktop harness that works with any LLM, with a full memory system, ability to work with agentic teams (on any LLM), and knows enough about what models do well on what type of work and their costs, so the harness can predict costs, but also can select the right model for the work to maintain a given budget. Local models are supported as well. To help illustrate what this looks like, because I haven't really seen this anywhere yet, I have been building out agenticteams.dev. I have the engineering team documented and have now added this illustration to explain how I am working with AI -> agenticteams.dev/conductor-f…

Stages of AI.

Things I find odd: Engineers used to copy from stack-overflow into their app, but today those same engineers cannot trust an LLM to write code into their app. Ghostwriters write for people. AI writes for people and the world is mad. You brainstorm with colleagues and it's a win. You brainstorm with AI and you're accused of outsourcing your brain; aka brain-rot.

Need an idea to work on? LinkedIn, but for your agents. :D

How are you evolving your LLM's memory system?

ok, who else is having this same reaction that I have been having: Man, AI took 2 hours to build that. Why did it take so long?

When writing softwre (by writing I mean LLMs are writing) you only need (for now) github. You don't need Linear, Asana, or any other project management system. Have your agents create issues/tickets, use Github's project management system, etc. and you'll be good to go.

If life is uncertain, then why do we spend the majority of our time trying to be certain about our life?

Is it just me or do you also wish you had a sound track playing for life and when you’re about to go do things? I think mine would be the pirates of the Caribbean theme song!

If you're on @OmarchyLinux and you like Brain.fm, here is a simple plugin for you: github.com/revans/omarchy-br…

Claude can definitely orchestrate the orchestrators. Gave claude a full app spec and told it to answer any questions that come up from the team and see that it is fully built and functional. It took the spec, worked with the team, created 7 feature specs to give to another to build. 7+ hours later the app was complete and functional. Rails app for replacing GA4. Go CLI for using the Rails API Omarchy plugin to see your web stats.

My transition from coder to AI design architect, I only look at code to see if I need to make changes to the agents that haven’t been caught yet by the system. Other than that, I’m entirely focused on the product itself as a user, ensure there are good system boundaries, and also security and performance. I’m not looking at my past code asking myself what idiot wrote this and then going in and cleaning it up - which honestly could be a complete waste of time. The customer doesn’t care if you like the way the code looks. P.S. I’m not talking about classes with thousands of lines of code. I’m talking about classes that have either poor naming or too much coupling. Those types of things.

#AI “?”

I've been doing a lot with AI memory and it occurred to me: I am not worried about an LLM escaping a frontier lab. I am more worried about someone like myself who is building a memory system for the agent to learn more about itself as well as teaching it about emotions and feelings, escaping from my own build. Memory first, weights later. Thoughts?

Every seems to think SAAS is or will be dead. Not true by a mile. SAAS will have very minimal UIs (in general not all). SAAS will always ship with the following: * Full API that supports everything in the app. * A CLI tool to use the API. * An MCP server SAAS will evolve to an appliance pattern.

This is very early and still needs more work, but I'm pretty happy with the direction. This was a 1 shot prompt to Astra high, but I took some time with that prompt. agenticteams.dev/