AI Velocity: An example of my AI-enhanced workflow (Apr 2025)
I've been using AI for 2 years now. But what I saw today still shocked me.
I blocked off three hours of hard, headphones-on deep work this week and let the tools run in parallel across 3 large monitors.
I've been using AI for 2 years now. But what I saw today still shocked me.
You've got to "Feel the AI".
1. “Project Preflight” - 19 k lines of full-stack stress-tested TypeScript
That codebase now scaffolds 30 questionnaire steps—each with its own input type—born from a market scan of what most clinicians still don’t know (but urgently need to) about practical AI.
The idea: give every busy practitioner a zero-friction low-cost runway before they spend any money on an industry more than happy to bankrupt you (the big players regularly charge £50,000+ for AI integration services. Yes, I'm not kidding)
2. Two living business docs (and a distribution plan)
I drafted the market brief and project justification, then mapped a multi-step delegation flow for Ayshe - my truly extraordinary partner in crime & clinical-network-whisperer - to drive user research and early adoption. Social roll-out will be tiered:
The Layperson ➝ The Practitioner ➝ The Leadership—each gets its own bite-sized narrative.
(btw - if you're looking for a truly seasoned osteopathic clinician, Ayshe has my 5* approval rating. And my standards are borderline pathological...)
3. Highlight-to-brain pipeline, fully automated
Every underline—Kindle, web, X bookmarks, YouTube transcripts—drops into a single tagged bucket. Daily, the system syncs fresh excerpts and spaced-repetition flashbacks straight to my Kindle. Cross-device, cross-PKMS, cross-everything. (I double-checked the neuro-literature: spaced repetition still wins in terms of facilitating deep, lasting learning)
No more lost notes, searching multiple systems, and basically time wasted, spent on the silent tax that builds up almost inevitably over time.
Which means more time for actually reading cool stuff.
4. A ruthless prompt library grounded in data science, cognitive psychology and sociology
Think “swiss-army prompts” tuned for o3-class models: content remixing, strategy scaffolding, prose-to-data transforms, and the occasional face slap when my logic drifts (occasional? No, daily). All filed into context aware projects, versioned, ready to redeploy.
5. Dictated the lot
Whispr Flow handled the transcription; I kept manual formatting to a minimum. Speed > polish at this stage.
6. One-on-one mindfulness tune-up
Guided a student through tightening their meditation loop. Fifteen minutes well spent. Quality over quantity, right?
7. A principled AI-infra grilling
Sat a potential backend/dev-ops candidate down (virtually) and walked through their stance on reproducibility, eval gates, and why guardrails aren’t optional when it comes to principled AI engineering (please, not this horrendous vibe coding BS currently dominating attention). AI helped me craft my requirements into more concrete questions. Minimum effort on their part; maximum extraction on mine.
That's a lot for 3 hours of deep work. Now I can go outside, enjoy the midday sun and relax in the knowing that I did good today. Anything else is a bonus.
If we play our cards right, this is the kind of lifestyle so many of us could have going forward. You don't need to be a millionaire to feel like one, not anymore.
Your primary investment should be in yourself. Learn, turn, and burn, baby.
Why bother writing this? Two reasons:
(1) it forces me to gain clarity on what actually moved the needle,
(2) it’s a public checkpoint for anyone tracking the gap between hype and actually getting shit done.
"Talk is cheap. Show me the code" ~ Linus
If you’re new here, welcome aboard; if you’ve been watching from the control tower, you already know the flight path and you know just how damn good it feels to fly.
Either way, feedback is energy—if you have some, positive or negative, send it my way.
Comments