In search of a life beyond traditional norms, exploring, learning, and building with AI.
South Africa
Joined May 2024
- Tweets242
- Following115
- Followers6
- Likes187
You have infinite intelligence.
And that's before we have even counted the intelligence in your head.
"The personal computer is a bicycle for the mind" – Steve Jobs
If the computer is the bicycle, your personal AI Agent is the jet engine strapped to the back.
Hold on and build everything you can think of.
The way we think of all people professionally will change in a world with AI.
People will go from "doers of tasks" to "owners of outcomes".
How does owning your outcome look for you?
Even if the AI bubble bursts, it won’t break the systems we’ve already built → because the tech is real, and the work is already done.
One of my favourite things about working with LLMs is how they teach me things I didn’t know as we work together on the task.
Will we ever not need a framework?
I used to think the model would do it all. Now I see: frameworks evolve.
We started with prompt → call → repeat. Now we build agent harnesses.
So yes… frameworks will always be there. They just won’t look like they used to.
Want to run your own Gen AI locally?
I’ve been using Open WebUI → hands down the best open-source chat UI for LLMs.
Plug in cloud or local models; same interface, no fuss.
Great for tinkering solo or exploring private AI. No coding needed.
DM if you’re curious
You should be proud of what you’re building.
Not just the outcome, but the courage it takes to start and keep going.
Share this with the first person you thought of.
They might need to hear it too. 💡
Today I was reminded → when coding with AI, do one thing at a time.
I was building a complex program with my local AI.
Got stuck.
Then realized: I wasn’t breaking it down.
So I wrote each step like a recipe.
Gave them one by one. AI got it. Program worked.
One reason enterprise AI fails?
The point-solution trap → fixing one problem without connecting the dots.
When done right, local AI becomes part of your org’s digital nervous system; linking teams, surfacing insights, and empowering everyone, not just data teams.
First we build our private, local AI → Then we work with it on all future projects.
What projects do you want your AI to help with?
Thinking about agent memory?
3 key ideas:
~ Knows you → tone, style, interests.
~ Learn from mistakes → don’t repeat errors.
~ Remembers the past → use last weeks chats.
And do it all fast → no lag.
Your local AI should remember, adapt, and respond… without slowing down.
Spent time reading LangChain V1.0 docs today.
The create_agent + Middleware is a game-changer:
✅ Controls context length
✅ Routes tools correctly
✅ Removes old results
Just simpler, reliable agents.
Great work, @langchain !
Docs: docs.langchain.com
Today, you can run your own LLM agent locally on consumer hardware.
Chat privately over your docs.
Run AI-powered Python notebooks.
Code locally.
Automate workflows.
Even talk to it → voice works now, and fast.
You don’t need cloud. Don’t need an API. Just your machine.
🚀 AI Agents vs. Workflows: Not the same.
Workflows = step-by-step, predictable.
Agents = self-directing, adaptive.
Use workflows for reliability.
Use agents for flexibility.
anthropic.com/engineering/bu…
This isn’t about replacing humans.
It’s about enhancing them.
I want an AI that works independently → but keeps me in the loop. Asking questions. Proposing ideas. Making me better.
It’s a teammate. One that learns with me. That builds alongside me.
Long-term. Private. Local.
Imagine this:
You design a mechanical system. Your AI pulls from your notes, the web, past projects.
It simulates. It builds. It tests.
When the prototype fails → it shows you why.
No guesswork. No blind automation. Just logic. Partnership.
What if your AI didn’t just answer → it builds with you?
I’m building a local AI tailored me.
It runs on my hardware. It sees my world.
When I design something; a machine, a system; it pulls from the web and my archives.
I guide. It generates. We build.