Building at the frontier of agentic AI
We build AI products and publish what we learn along the way. Research, tools, and open knowledge from the cutting edge.
What we're exploring
Our research focus
We build products and conduct research across several areas of agentic AI. Here's what we're working on.
Autonomous Agents
Agents that reason, plan, and execute multi-step tasks with minimal human oversight. We study how to make them reliable enough for production.
Multi-Agent Orchestration
Coordinating teams of specialized AI agents that collaborate, debate, and synthesize — the architecture behind Kapwa's Symphony Mode.
Long-Horizon Research
Deep research agents that work over days and weeks, building on their own findings to produce continuously evolving analysis.
Applied AI Engineering
Turning research into shipped products. Streaming architectures, semantic memory, tool use, and the engineering that makes AI systems work.
See agentic AI analysis
in action
Describe any work or business scenario and get a complete build-vs-buy analysis — custom agent architectures, off-the-shelf product recommendations, and a practical comparison to help you decide.
Our Products
Kapwa
Our flagship AI product — an advisor platform where users select from hundreds of specialized personas — historical figures, domain experts, and fictional strategists — for multi-perspective conversations powered by ensemble AI orchestration.
What's next
Long-horizon deep research
We're building a research agent that doesn't stop after one answer. It produces a report, then continues working — running deeper analysis, finding new connections, and updating its findings daily. Research that compounds over time.
Continuous research reports
Imagine a research report that updates itself. The agent performs an initial deep dive, delivers findings, then keeps working in the background — running increasingly complex analyses that build on previous results. Each day, the report gets deeper and more nuanced.
Reading List
What we're reading
Weekly Gen AI headlines for builders, plus the papers that define the field.
Anthropic restricts OpenClaw usage via new Claude Code pricing tier
Anthropic is introducing extra fees for Claude Code subscribers using third-party tools like OpenClaw. This signals a shift toward vertical integration and tighter control over the developer ecosystem
3D-Layout-R1: Structured Reasoning for Language-Instructed Spatial Editing
This paper addresses the limitation of large language and vision-language models in maintaining spatial consistency during fine-grained visual editing by introducing a structured reasoning framework that operates over scene graphs. By reformulating text-conditioned spatial editing as explicit graph reasoning rather than end-to-end generation, the method enables precise manipulation of object layouts through natural language instructions while preserving geometric coherence. The work establishes structured scene-graph reasoning as a necessary intermediate representation for bridging high-level linguistic commands with geometrically consistent spatial editing in 3D environments.
Agentic Reasoning for Large Language Models
Comprehensive survey organizing agentic reasoning into three layers: foundational (planning, tool-use, search), self-evolving (adaptation through feedback and memory), and collective (multi-agent coordination and role specialization). Bridges in-context reasoning with post-training approaches across science, robotics, healthcare, and mathematics applications. Accompanied by an actively maintained Awesome-Agentic-Reasoning GitHub repository.
Learn
What we've learned
Notes, frameworks, and explanations from our research and product work. Written to be useful, not to sell.
What is AI (Without the Jargon)
A plain-English explanation of AI, machine learning, and large language models — no jargon, no hype.
Read moreWhat are AI Agents
From chatbots to autonomous agents — what makes an agent different and why it matters.
Read moreThe Agentic AI Framework
A structured approach to identifying where AI agents create the most value in any operation.
Read more