Notes about my work, projects, training, and life.
Build note: I’ve been tightening the system behind how I publish on JimLunsford.com.
The goal is to make drafts, review passes, metadata, WordPress formatting, and final publishing follow the same repeatable process instead of relying on me remembering every little step.
Good publishing systems should reduce friction without lowering the standard.
Day off. Building a fence in the backyard.
Stopped for lunch and Blu climbed into my lap. Apparently she has issued a stop-work order.
Fence construction will resume when management approves my release.
Recovery gets stronger when you stop needing permission to do what you already know is right.
You also stop needing praise every time you follow through.
The standard should hold whether anyone approves, notices, or congratulates you.
Do the right thing because it is yours to do.
No permission.
No applause required.
https://jimlunsford.com/recovery-standard-no-permission-or-praise/
Something changed in the way I build software with AI.
The repository stopped being something I had to fully explain at the start of every session.
The code, documentation, architecture, decisions, and history became context the AI could work from.
That changes the workflow.
The repository is no longer just where the software lives.
It becomes part of the AI’s memory of the project.
https://jimlunsford.com/when-my-repository-became-part-of-the-ais-memory/
One thing I have learned in corrections is that respect works better than trying to prove you are in charge.
The uniform already tells people what authority you have. You do not need to manufacture more of it every time somebody challenges you.
Some people will test you. Some will lie to you. Some will try to manipulate you. You still get to decide how you carry yourself.
You can hold the line without turning every interaction into a power struggle.
Real authority does not need constant theater.
I’ve been pushing Only Empowerment toward its first release candidate.
It’s a privacy-first collection of practical tools for making decisions, setting standards, resetting after a miss, and turning intentions into action. No AI making choices for you. No scoring or gamification.
The goal is simple: build tools that help people think and act for themselves.
Turning 50 changed how I think about time.
I still want to build, work, run, write, and do hard things.
I just pay more attention now to what each of those things costs.
Time is not abstract anymore.
I want to spend it on things that are worth the piece of my life they take.
I do not like systems that require permanent dependency.
That applies to recovery, technology, work, and just about everything else.
Support is useful. Tools are useful. Structure is useful.
But the goal should be greater capability, not needing the system forever just to stand.
Bonumark Stream v0.8.2 is out.
This release is less about adding shiny features and more about making the system behave correctly under real use.
Unified Like and Comment counts. Better federated replies. Safer composer recovery. Stronger accessibility. Cleaner ActivityPub administration. Better media handling. More reliability work underneath it all.
Here is what changed and why it matters.
I’ve been testing GPT-6 Sol in ChatGPT Work across real development projects, and I keep seeing more execution mistakes than I do with Astra.
Missed instructions. Incomplete gates. Context that gets read but does not always survive the full execution.
I went looking to see if it was just me. It isn’t. Other users are reporting similar problems with context handling and long-running Work/Codex sessions.
That does not mean Sol is universally worse. It does mean I’m changing how I use it.
For technical project work, I’m back to Astra first, usually at the lowest reasoning level that can reliably finish the job.
Sol still has a place.
It just is not my default for technical Work anymore.
