It has been almost exactly three years since we started out with Locus Extension.
We have been pretty quiet publicly during that time. Not because this is new for us, but because we have been busy building through the hard part of doing agentic AI right.
Easy to talk about. Much harder to get right.
A lot of the market is now running into the same thing. Most teams do not have a problem getting started with AI, chatbots, and agents. The harder part is execution. Then scaling. Then making the work actually hold up under real conditions.
It is not that the models cannot do useful things. They can. But like any tool, a lot depends on who is using it, in what situation, and against what constraints. Quality varies wildly. Getting from a good answer in a controlled setting to work you can actually rely on is a very different problem.
Real knowledge work comes with constraints. Messy source material. Permissions. Review. Traceability. The need to hand someone something they can actually use, not just something that sounds plausible.
That is the gap we have been building in for a while.
Today we are launching Ikentic, our platform for Knowledge Work Agents and agentic applications, and the new OmniArcs website.
Ikentic is the product layer. OmniArcs is the studio and delivery layer around it.
Execution is where the work starts
This is the part we kept running into. Once you move past the demo, the problem changes. The question is no longer whether the model can produce something useful. It usually can.
The better question is whether that useful output can become dependable work.
That is where things get harder. The system has to know what job it is doing. It has to know what knowledge is in scope. It has to stay within the right boundaries. It has to use tools in a controlled way. It has to produce something someone can review, trust, and use.
That sounds obvious, but it is where a lot of agentic work breaks down.
The work around the AI is often under-specified. The prompt is doing too much. The workflow is too loose. The source material is messy. The review path is unclear. The output sounds right, but it is hard to tell what it is based on or whether it should be trusted.
That is not a model problem by itself.
It is an execution problem.
And that is the layer we have been building toward.
The Integrated Knowledge Environment (IKE)
That is what IKE is meant to address.
IKE stands for Integrated Knowledge Environment, which is a fairly literal name. The point is not to add another chat interface on top of scattered information. The point is to give agentic work an environment where the knowledge, context, tools, roles, and review paths are part of the system.
That matters because real knowledge work is rarely just one prompt and one answer.
It depends on what the agent knows, what it is allowed to use, what job it is supposed to do, and what evidence it can bring back with the work. If those pieces are not integrated, the agent may still produce something that sounds good, but it is hard to know whether the work is scoped, grounded, or usable.
That is the gap IKE is built around.
Knowledge Collections give the work a grounded base. Personas define role, behavior, and context. Skills connect the system to actual workflows and tools. Lineage and review make the output easier to inspect, trust, and improve over time.
The goal is not to make agents feel more impressive in a demo.
The goal is to make knowledge work more dependable in practice.
Not magic. Not “set it loose and hope.”
A more deliberate environment for defining what the agent is supposed to do, what knowledge it can use, what tools it can touch, and how the work gets reviewed.
How we got here
The path here did not start with the current AI wave. Before Locus and before OmniArcs, the team had already spent years building around data engineering, cloud, knowledge systems, and product delivery in our heritage firm. OmniArcs was the new venture and product studio around that work.
Locus Extension was one of the first focused product expressions of that direction.
It started with a practical problem: people were doing more and more of their knowledge work in the browser, across articles, documents, research trails, PDFs, internal pages, and scattered context that was hard to find again when it mattered.
The first job was to make that context easier to capture, search, and reuse. That was useful. It still is.
But the more we built, the clearer the next problem became. Finding information is only part of the work. Summarizing it is only part of the work. Even chatting with it is only part of the work. At some point, the question becomes: can this knowledge actually help do the work?
That is a different problem.
It means the system needs to understand the role it is playing, the knowledge it is allowed to use, the tools it can touch, the constraints around the work, and the evidence needed for someone to review the result.
Locus gave us the first-mile capture and context layer. IKE is the broader Integrated Knowledge Environment we have been building from that foundation. It brings together Collections, Personas, Skills, workflow context, evidence, and the controls needed to make agentic work more dependable.
OmniArcs is the venture and product studio behind it. It is where we shape use cases, build with teams, and keep pressure-testing the difference between something that works in a demo and something that can actually be used in practice.
Ikentic is the product and platform layer coming out of it. IKE is the Integrated Knowledge Environment underneath it.
That is the progression from Locus to IKE.
What is live now
What is live now is Ikentic, along with a better OmniArcs site around it.
On Ikentic, you can see the core platform ideas in public: what IKE is, the personas you can build, the knowledge collections that keep work grounded, and the direction around skills, lineage, deployment, ADK, Knowledge Work Agents, and agentic applications.
On OmniArcs, you can see the broader frame around it: the studio behind the products, the move from data platforms to knowledge platforms, and the way we think teams should move from prototype energy to governed systems that can hold up in production.
Where this goes
We think the next wave here will be defined less by who can stage the best demo and more by who can make knowledge work actually run.
That means systems that are grounded, inspectable, governable, and useful under real conditions. It also means agentic applications that can act without becoming opaque or unstable.
More broadly, we think the center of gravity is moving from data platforms to knowledge platforms.
The next important systems will not just store information or answer questions. They will help organizations turn knowledge into work that can be delegated, reviewed, improved, and trusted.
That is what we are building toward.
We will be sharing more of what we are seeing from this layer. What breaks. What matters. What gets overstated. What teams underestimate. And what it actually takes to make agentic AI useful.
If you want the platform view, start with What is IKE. If you want the studio view, start with About OmniArcs. If you want the deeper framing, read the Ikentic whitepaper.
