November brought new flagship models from Anthropic and OpenAI — and they were not the main event. Microsoft told Ignite it already counts more than 50 million agents under one control plane, and IBM and Microsoft launched a platform whose first promise is governing agents. The question for a business has moved from “what can an agent do?” to “who answers for it?”
Insights
Clear thinking, better systems.
Straightforward notes on software, AI, product design, mobile, and search — written for people who need practical decisions, not noise.
All articles
Agents That Ask Before They Act: What October’s Releases Tell Us About the Next Year of Software
Anthropic’s Claude Duet watches your connected apps and checks with you before it sends, pays, signs or deletes. GPT-6 Pro now splits one task across several agents. Gemini 4 Argon went public after a defenders-only launch. The direction is the same from every lab — and it should change how businesses design software.
AI
Always-On Agents Have Arrived. The Interesting Part Is the Permission Model.
September brought a new frontier model from nearly every lab, sharp price cuts, and the first mainstream always-on agents — OpenAI’s dots in ChatGPT and Microsoft’s Autopilot in Copilot. Underneath the headlines, every one of them ships the same thing: a policy for what an agent may do alone, what it must ask about, and what it may never touch.
Agents
The Price of Intelligence Keeps Halving. Your AI Budget Should Not Be About Tokens.
In August Google shipped Gemini 3.7 Flash three weeks after 3.6 Flash, at half the price, while computer-use and browser agents went generally available. For a business planning an AI feature, the model is now the cheapest line on the estimate — and that changes how you should scope the work.
AI & Cloud
2D vs 3D Web: When to Use Each, What It Costs, and How It Performs
WebGL runs in 97% of browsers, the three.js core alone weighs a third of the median page’s JavaScript, and nobody has published a fair benchmark of 3D sites against flat ones. As a studio that builds both, we lay out the measurements that exist, the ones that do not, and how we decide.
Web
Kimi K3 and the Rise of Open Frontier AI: What It Means for Your Business
Moonshot AI just released Kimi K3 — a 2.8-trillion-parameter open-weight model that trades wins with the top proprietary systems. We look past the benchmark race at what open frontier AI actually changes for the businesses building on it.
AI
OCR and Document Automation: Turning Paperwork Into Workflow
OCR automation is useful when documents, invoices, forms, IDs, receipts, and records need to become structured data with review queues and audit trails.
Applied AI
AI Automation for Malaysian SMEs: Where to Start Safely
AI automation should begin with low-risk, high-volume work: classification, summarization, document review, routing, recommendations, and human-in-the-loop decisions.
AI
Workflow Automation Examples for Malaysian Businesses
Practical workflow automation examples for approvals, claims, scheduling, document review, notifications, reporting, and operations dashboards.
Automation
How Much Does Mobile App Development Cost in Malaysia?
Mobile app cost depends on user accounts, design quality, notifications, offline behaviour, admin tools, integrations, and store launch support. Here is how to scope it properly.
Mobile
Web Application Development in Malaysia: What Businesses Should Plan First
Malaysian establishments sell RM1.29 trillion online, 99% of internet users are on a phone, and yet most software projects still miss their brief. We set out the local numbers, the failure data and the planning decisions that separate the systems that ship from the ones that stall.
Web
Custom Software vs SaaS: Which Should Malaysian SMEs Choose?
A decision guide for Malaysian SMEs weighing custom software against off-the-shelf SaaS tools for operations, approvals, reporting, and workflow automation.
Strategy
How Much Does Custom Software Development Cost in Malaysia?
A practical guide to the factors that shape custom software development cost in Malaysia, from workflow complexity and integrations to dashboards, permissions, and support.
Software
Integrating AI Into Your Product Without Rebuilding Everything
You do not need to rewrite your stack to use modern AI. Here is how we add LLM capabilities to existing systems cleanly, without blowing up timelines or trust.
AI & Cloud
React Native vs Flutter: A Production Perspective
Framework debates rarely survive first contact with a real project. This is our honest assessment after shipping production apps in both — what we reach for, and why.
Mobile
Building for Scale: What Most Teams Get Wrong About Architecture
A system that works for 100 users will crack under 100,000. We share the architectural decisions that prevent costly rewrites — and the patterns we reach for on every project.
Engineering
How Technical SEO Became the Most Underrated Growth Channel
Fewer than half of mobile sites pass Core Web Vitals, a tenth of a second is worth 8% in retail conversions, and AI Overviews now take more than half the clicks from the top result. We read the measurements and set out what technical SEO is actually worth in 2026.
SEO
Choose how to get started
Build with us
A system engineered around how you work.
Scoped in a proposal, designed as a prototype, delivered in milestones you can review.
Start a projectSee the work
Systems already running in production.
Marketplaces, operations portals and mobile apps, with the case studies behind them.
View our work