When you keep losing the thread
How I Stopped Losing Context With AI
The practical one: how Obsidian helps me pick up a project after an interruption.
Read on Substack ↗Ideas → experiments → useful things
Writing
An index of work published on Substack. Each entry links to the full original article there.
New here?
When you keep losing the thread
The practical one: how Obsidian helps me pick up a project after an interruption.
Read on Substack ↗Why this work matters to me
The personal one: Purple, AI and finding support that lets me keep being myself.
Read on Substack ↗When the software feels like too much
The starting point: what finally made Obsidian usable after I had repeatedly given up.
Read on Substack ↗Explore the archive
Search the useful bits here, then read the full original article on Substack.
Showing 49 posts
AI & systems · 2026-11-17
A countdown for the space station grew into OMANOVA. How curiosity, AI help, practical checks and having an opinion about the result changed what I thought I could build—and how to start with one small idea of your own.
AI & systems · 2026-11-12
A free starter-vault alpha for finding your place after time away. Why I built it, what the first iMac test exposed, how optional AI guidance works, and the beginner feedback I need next.
AI & systems · 2026-11-10
I tested the same SignalTube install guide with Codex alone and a Codex–Grok review in OpenBot. What the second reviewer caught, what still needed correcting, and a companion PDF for trying one review yourself.
AI & systems · 2026-10-27
I introduced an AI assistant to my notes. Then it gave me a job I had already finished.
AI & systems · 2026-10-20
Building a touchscreen companion for my Linux desktop, one awkward step at a time.
AI & systems · 2026-10-13
The articles, small tools and conversations that gave other people something concrete to respond to.
AI & systems · 2026-10-06
My first few days with ChatGPT dots and Space, alongside the second brain I already rely on.
AI & systems · 2026-09-29
How I brief AI, challenge it, get a second opinion and keep a way back when it gets things wrong. Practical lessons from writing, Linux projects and everyday use.
Parkinson’s & personal · 2026-09-25
What has helped, what has not, and what I understand differently now: learning to ask better questions, recover context, question the answers and recognise when another system would ask too much back.
AI & systems · 2026-09-22
Testing whether preferences, saved procedures and useful context survive a fresh session or a change of model. Early Hermes results, a bounded workflow test and the distinction between remembering a boundary and enforcing it.
Knowledge & systems · 2026-09-15
An old note can preserve what I thought without remaining the right advice for today. How I use AI to compare earlier claims with later decisions, keep the history and check the evidence before acting.
Lab Notes · 2026-09-08
A reversible portable-Linux experiment became my everyday computer. The hardware round trip, local-AI results, rough edges and a practical guide to planning your own move.
AI & systems · 2026-09-04
Not a robot butler. More a set of useful hands that do not get to touch the steering wheel.
Knowledge & systems · 2026-09-01
How I use Obsidian as external context for AI, so projects survive a model change, a week away, or a brain having a bad day without re-explaining it.
Parkinson’s & personal · 2026-08-28
Purple helps me manage the physical world; AI helps make the cognitive one more manageable. Support is not surrender, and the final judgement stays human.
AI & systems · 2026-08-25
How I built a small static website with Astro and Cloudflare for about $12 in new costs—without renting a website builder or pretending it was free.
AI & systems · 2026-08-21
Useful is not the same as right. Why I question confident AI answers, ask for evidence, and keep the final decision human.
AI & systems · 2026-08-18
I tested AI on small, private vault bundles, tracked the cost, and tried Qwen as a source-gated drafting option, not a substitute for authorship.
AI & systems · 2026-08-11
A field note from a five-set local test: useful connections, one correct refusal, and no permission to touch the real vault.
AI & systems · 2026-08-04
Let the system remember the cycle. Keep the part that becomes your message with you.
Knowledge & systems · 2026-07-31
The interesting part is not that an AI can reach my tools. It is that it can do so without being mistaken for the person in charge.
Parkinson’s & personal · 2026-07-28
The fix took ten seconds. Remembering it was the expensive part.
AI & systems · 2026-07-21
The simple rule that stops a useful assistant taking orders from every document it reads.
Parkinson’s & personal · 2026-07-17
Obsidian Dashboard: A Calm Front Door to Your Vault
Knowledge & systems · 2026-07-16
NotebookLM turned my article about an AI-maintained Obsidian wiki into a 24-minute conversation that understood the point better than it had any right to.
AI & systems · 2026-07-14
What a narrow write lane, explicit guardrails and human review look like when the vault is part of your external memory.
Parkinson’s & personal · 2026-07-07
The pipeline still works. The missing piece was proving its own claims.
AI & systems · 2026-07-01
Two AIs, one shared memory — and no more pasting the same paragraph at 3am
Parkinson’s & personal · 2026-06-28
A field report from a brain that was already losing.
AI & systems · 2026-06-23
What I built when growth started outrunning my brain.
Parkinson’s & personal · 2026-06-18
I am not building this because I am clever. I am building this because I forget things.
Parkinson’s & personal · 2026-06-15
The productivity system that doesn't punish me for being ill.
AI & systems · 2026-06-10
The day I discovered my AI instructions were the problem
Parkinson’s & personal · 2026-06-08
Everyone says niche down. I calculated which version of me was most marketable. Then nine people showed up and the question stopped mattering.
Parkinson’s & personal · 2026-06-04
Parkinson's left the ideas intact but made the follow-through painfully hard. AI became a bridge between the thought and the finished document.
Parkinson’s & personal · 2026-06-02
97 saved articles. 74 wiki pages. Zero manual linking. One degenerating brain.
Parkinson’s & personal · 2026-05-29
Why "challenge me" are the two most powerful words you can give an AI
Knowledge & systems · 2026-05-26
I quit Obsidian somewhere between ten and fifteen times. I lost count. This is the story of the session that finally made it stick — and why the breakthrough wasn't about Obsidian at all.
Parkinson’s & personal · 2026-05-22
The midnight thought that wouldn't let me sleep — and why the answer matters more than you think
AI & systems · 2026-05-21
Everyone's talking about agents. Most people are describing a chatbot with a plugin. Here's what's actually going on.
Knowledge & systems · 2026-05-20
Most people have a prompt graveyard. Here's how I built a vault that hunts for them automatically — and why it took a brain that can't afford to forget things to finally solve the problem properly.
Parkinson’s & personal · 2026-05-19
A weekly look at my Obsidian vault, AI experiments, projects and progress.
AI & systems · 2026-05-17
The AI tools I was actually using, and how they fitted into everyday life.
Parkinson’s & personal · 2026-05-13
How I used two AI models and a note-taking app to support a memory that does not reliably hold on to things.
Parkinson’s & personal · 2026-05-12
A weekly review, an honest account, and a system that doesn't require you to be fixed first.
AI & systems · 2026-05-10
How my use of AI grew into a collection of digital assistants, and what I learnt along the way.
Knowledge & systems · 2026-05-10
From productivity-app chaos to an AI-assisted second brain: my experience of finding a system I could work with.
Parkinson’s & personal · 2026-05-08
I started writing this book because I couldn’t stop thinking about it.
Parkinson’s & personal · 2026-05-07
Let me be upfront with you: I didn’t plan to start writing.
No matching writing here yet. Try a different search, subject or tag.