We Don't Need Bigger AI. We Need Smarter AI Infrastructure.
From a systems-build standpoint, our first deliverable is already solid: a matured Detached System running on a LAMP-based architecture. The rule is simple - not every task needs an LLM. Validation, business rules, reconciliation, workflow control, repetitive logic and a lot of operational decisions can be handled by deterministic code outside the model.
Over the last six months, even though the full platform is still a work in progress, we have shipped around 250 Detached Systems. Total development cost so far: less than USD200, including dead-ends, retries, failed experiments and the mistakes you expect when you build fast. That is practical proof that useful business systems do not always need big budgets.
The next layer we are wiring up is a modded AI agent with Smart Routing capability. Instead of defaulting every request to the biggest model in the stack, the agent is being built to continuously pick the smallest, lowest-cost LLM that can still handle the task reliably.
The routing logic is direct: simple task → small model, difficult task → stronger model, no intelligence required → Detached System. The point is not just to cut token spend. It is to make AI infrastructure economically sustainable by putting intelligence only where the problem actually needs it.
The third layer is our own set of specialised LLM models, tuned to work alongside our Detached Systems. We are currently testing 7 open-source LLM models as the base for this. Hugging Face will be part of our technology ecosystem, giving us access to open-source models, datasets, training tools and deployment infrastructure.
But these three layers are not the end goal. They are stepping stones toward something larger: making system development accessible to ordinary SME owners - businesses without unlimited funding, GPU farms, IT teams or technical depth.
Eventually, even a makcik selling pisang goreng by the roadside should be able to become the system developer for her own business. She should just be able to say, "Manage my stock, calculate my daily profit, monitor ingredient costs, remember my regular customers and tell me when I need to buy more bananas," and the agent should assemble the system behind that.
That is the roadmap: Detached System → Smart Routing → Lowest Suitable LLM → Specialised Own LLM → AI-Assisted System Development for Everyone. Minimum cost is the immediate target. Free is the dream. We are not chasing the biggest AI. We are making AI and system development small, cheap and simple enough that the smallest SME can actually build with it.
#AINNA #NeuralOps #HuggingFace #OpenSourceAI #LLM #AIInfrastructure #SmartRouting #DetachedSystem #SME #SystemDevelopment #Automation
