Field notes from the people building trusted AI. Short, sharp dispatches on the intersection of technology and humanity.
Physical AI is moving robots from the lab into everyday workplaces. We look at how AI helps them move and see, and at the patrol robot we triggered from WhatsApp in our own first experiment.
Read dispatch →Physical AI is moving robots from the lab into everyday workplaces. We look at how AI helps them move and see, and at the patrol robot we triggered from WhatsApp in our own first experiment.
Digitising your documents was always step one. The businesses actually seeing returns are the ones who built workflows on top of that data, and are now pushing one storey higher, to AI transformation.
Singapore's courts have been digital since the 1990s, with AI-powered translation and case summarisation already live. A recent defamation case still ran up a S$35,000 printing bill. Here is why, and what closing that gap would actually take.
What if every business, every service, every policy had a rehearsal space, a working simulation of the real world, before it ever touched a real customer or a real citizen? This is not a finished idea. It is one we have been sitting with, and Singapore may already be proving parts of it work.
For decades, depth won. A degree, a specialism, three or four years spent going narrow and deep. AI is quietly rewriting that trade. Here is what the rise of the generalist actually means for how organisations should hire, train, and structure their teams.
The typical RFP asks a vendor to promise a solution nobody will see working until months after the contract is signed. Using Trusted ActionVerse, we build the actual thing before the pitch even happens, and we think that changes what customers and vendors should expect from each other.
The AI jobs apocalypse that was promised has not shown up in the data. What is showing up instead is something better, a chance to strip away the busywork and get back to the part of the job that made people want it in the first place.
As AI decouples cognition from human labour, data and computing capacity are becoming the new sources of national wealth. Here is what that means for geopolitics, for Singapore, and for the small businesses being left behind by the price of admission.
On 5 August, MDDI told Parliament that agentic AI doesn't get its own rulebook. It falls back on the same principle Singapore has held since January: a named person stays accountable for what the agent does.
We spent Friday at ITE, talking to single-man business owners and SME association leaders about the one thing in their day that hurts the most. What we heard reaffirmed why we build in Singapore, for Singapore, ahead of National Day.
Satya Nadella says companies that rely wholly on the major AI labs may not survive. He is right, and the reason is simpler than he made it sound. Every prompt you send is data you cannot get back.
A model is fixed. Its weights cannot be changed by the people using it. What you can control is the harness around it, the layer that decides what a model is allowed to do, and when it has to stop and wait for a human. In defence, that layer is not optional.
The framework for exponential organisations is clear: break down data silos, enrich what you collect, and let AI reason across all of it. Most companies have the data. They just cannot use it yet.
Gartner predicts 40% of agentic AI projects will fail by 2027. Microsoft's AI Red Team has mapped exactly why. Industry leaders are diagnosing the same root cause, consistently. The answer is not the model. It's never the model. It's what the model has to work with.
Singapore's total fertility rate hit 0.87 in 2025 — an all-time low. In a country built on national service, fewer births today means fewer soldiers tomorrow. The SAF must become smarter to compensate. But defence AI that depends on foreign servers is not a solution. It is a new vulnerability.
Singapore's Health Information Act now mandates every licensed clinic and hospital to contribute patient data to the National Electronic Health Record. The vision is right, the legislation is passed, and the deadline is 2027. The data problem, however, does not get solved by parliament alone.
Last week, hundreds of millions of people outside the United States lost access to the world's most advanced AI model overnight — not because anything broke, but because one government decided so. For a small nation whose only real resource is its people, that should be a wake-up call we cannot afford to sleep through.
Member of Parliament Yeo Wan Ling stood up in Parliament and told the House about a 25-year-old Singapore SME that took its people into AI as stakeholders, not passengers. That SME was TrustedHub.
Uber burned through its entire 2026 AI budget in four months. Tokenmaxxing is dead, the bills are landing, and a handful of US hyperscalers are smiling to the bank. Here's what Singapore enterprises should do before the same thing happens to them.
MAS Notice 314 makes life insurers keep records for the best part of 55 years. Warehousing a box costs $600; digitising it costs $140 — court-admissible and AI-ready. The maths is simple.
Most AI demos stop at the screen. Book a seat at the foundry and watch the whole operation run — paper to live agents, on your own governed data, in real time.
Frontier AI is hitting a wall that feels almost mundane for such a futuristic technology: it is running out of human-generated data to learn from. The scarce resource it needs next may already be sitting in your archive.
We handed a virtual insurer to student-built AI agents. Three findings on autonomy, unpredictability, and why governance is a first-principles design decision — not a checkbox.
Section 116A is the quiet reason TrustedHub records hold up in court — what "approved process" means, and why the burden of proof flips in your favour.
TechBytes is coming soon to your inbox. Drop your email with us now and you'll be the first to receive contemporary insights on the intersection between technology and humanity, every week.