29 May: self-observation notes
I needed to offload so fast today.
Before getting into the painful process of context-retrieving where I was yesterday, I had to capture what just came up.
Because that context-switch alone can make me forget what I’m thinking about now, even though I know I’ll need it later.
After I finished writing, I felt kind of relieved.
Today’s build thread:
- Upgrade items to handle carried items and their history.
- upgrade vee (thinking process)
- upgrade state building layer
- upgrade building workload function
- Upgrade validation and parsing,
- add replay formatter. (Constructing the user message + telegram html)
Note for future me:
Sometimes the first dev task is not coding.
It’s dumping the active mental stack before loading yesterday’s one.
It’s crazy that people still think the race is about bigger LLMs.
We’ve already compressed an incredible amount of human knowledge into these models.
Now what?
The next frontier isn’t more knowledge.
It’s continuity.
Agents that don’t just answer questions they remember.
Agents that understand you so well they can keep a conversation going with the same context, personality, and pace as if you never left.
But behind the scenes, they have operators that live in time.
They execute workflows, accumulate state, and continuously build context over days, months, and years.
This is where it gets even more interesting.
I think everyone will end up building their own AI architecture.
Your own “brain” that you can plug any LLM into.
The LLM becomes a component, not the product.
I’m talking about a future that’s 80% architecture and 20% LLM.
Persistent memory.
Structured and unstructured data.
State management.
Workflow orchestration.
Tool execution.
Long-term planning.
That’s where the real differentiation will come from.
The future isn’t AI that thinks.
It’s AI that continues.
@SukaruAnow@bridgemindai 😂 just wait.
All people u know today will be gone soon.
A lot of people are working in dark rooms to take the stage next.
AI will level the field for all of us.
@TrollColors Oh, It's only going to get more heretical over the next five years.
The first generation of truly powerful solo founders will operate with the efficiency that only large companies have today.
If you're building alone, don't optimize for looking like everyone else.
Learn from others, take inspiration where you find it, but never build a copy.
Your work should feel unmistakably yours.
Why do you think hosting Chinese open source models on your own GPUs is less trustworthy and secure than giving away your data to anthropic and being shut down access at any time because they think you are doing something they consider “not safe”?
I wish Claude Code or other CLIs could sometimes look for open-source projects for some elements of a project instead of spending tokens re-inventing the wheel.
Research, inspect, and decide yes or no.
“Open source is very dangerous” @DarioAmodei testifies in this stunning video. The real danger isn’t open models; it’s @AnthropicAI convincing Washington that only they can be trusted with AI. Open innovation beats regulatory capture [1/2 👇]