AI is an amplifier.
Not an equaliser.
It magnifies whoever is directing it. Hand it to someone who understands the problem and it moves fast. Hand it to someone who doesn't and it produces confident nonsense, just faster than before.
And there is no autonomous agent doing the directing for you yet. What exists today is workflow engineering, version two. I build with it, teach it, and I'm not selling you the version that doesn't exist.
Built for owners, not for developers.
Nineteen years across AI research, startup building and Fortune 500 strategy work went into deciding what actually matters for a business owner who needs this working by Friday, not a framework comparison.
A good fit
- You run a business and want AI doing real work in it, not sitting in a slide.
- You want the workflow, the prompts and the failure points, not the theory.
- You're comfortable being shown something that doesn't work yet, honestly.
Not a fit
- You want a certificate more than a working system.
- You want autonomous agents running your business unsupervised. Not built yet, anywhere.
- You want hype. Plenty of that elsewhere.
10M
Monthly pageviews at ShopperBoard, the platform he co-founded, scaled, and eventually sold.
Nineteen years in. Still shipping.
Numbers instead of adjectives, because adjectives are what everyone else in AI marketing is selling.
Projects that had to work, not just demo well.
Some are his own ventures. Some are client builds he can show the metrics for but not the source. Both kinds had to hold up under real usage.
nanogent.ai
Built to find out what workflow engineering actually looks like once it has to run unattended, not what it looks like on a roadmap slide.
LearnParrot.ai
A language-practice product used as the working proof for how he structures AI-assisted products end to end.
Read the case study ›Business+AI
Built for operators putting AI into real budgets and real headcount decisions, not pilot projects that never leave the pilot.
HelloBase.ai
Built so a small team's internal knowledge stays searchable and correct without someone maintaining a wiki nobody reads.
Small cohorts. Tools you leave using.
In-person sessions at the Hashmeta office in Singapore, four to seven hours, capped small so everyone leaves having actually built something, not watched a demo.
Four clusters. Written to be used.
Hub-and-spoke, not a scattered blog. Each cluster answers one kind of question a global, English-reading audience actually searches for.
5+ stages. Same argument, every time.
Amplifier, not equaliser, said to rooms that ranged from university faculty to enterprise marketing leads.