01
Read the source
Inspect what a tool does, its licence and limitations, and the problem it could help solve.
About · The studio and lab
An independent
Canadian
AI studio.
AgenticEdge is a creator-led studio for useful tools, original work and practical AI. The aim is to expand what people and communities can do, while earning a sustainable livelihood through work they value.
The studio · 01
Useful systems.
Room to explore.
We start with a person, a problem or an experience worth making. For a workflow, that means agreeing what better looks like and comparing quality, human effort and cost. For creative work, it means a coherent experience and feedback from the people who try it. Research supplies options; the result determines what is worth keeping.
Consulting connects technology research with feasibility, integration, training and scoped support. The applied lab compares local models, tests agent permissions and develops reusable procedures. Longer-term research informs that work without becoming an untested promise to a customer.
The ambition is a studio that delivers useful client work and develops original software, learning materials and creative projects. These projects can serve their own customers, and a useful product may become a business in its own right. Our current explorations include tools for clearer AI release decisions and an observable ant-world prototype. Applied research informs what we develop further. AgenticEdge is founder-led; as projects grow, the aim is to work with specialists who can help own delivery, product development and support.
The practice · 02
The practice
behind the work.
AgenticEdge combines customer-facing operations experience with self-directed AI and software development. The founder's practical route into technical work began with Python automation. Today we direct and review local AI experiments, evaluation tools and creative prototypes, using AI tools for bounded parts of the work.
The lab uses owned compute for experiments and workflows. In September 2026 we ran a coding assistant on a laptop over a phone hotspot, with the model running on our workstation back in the lab. We want to make useful parts of the work easier for other people to run and adapt.
Research into practice · 03
Research you can follow.
Decisions you can inspect.
A useful tool has to fit a real task. Our practice connects technical research, repeatable comparisons and clear operating instructions.
02
Test the workflow
Compare with the current method. Keep failures visible and count human checking, correction and operating cost.
03
Explain the result
Make the comparison useful to the person responsible: a recommendation, remaining questions and an agreed next step.
The lab supplies worked examples and lessons from failed comparisons. Each customer workflow still needs its own permitted inputs, checks and acceptance decision.
Workstation · 04
What the studio
runs on.
The primary workstation supports local inference and experiments. Each workload needs its own memory, speed and quality checks.
- GPU
- NVIDIA RTX 5090
- 32 GB VRAM
- CPU
- AMD Ryzen 9 9950X
- 16 cores · Zen 5
- Memory
- 96 GB DDR5
- Dual-channel
- Use
- Local AI experiments
- Model, context and workload checked per task
Client hardware is selected for the agreed workload and operating requirements. An assessment establishes the necessary capacity, where data will be stored and processed, and which external connections remain.
The workstation is also the lab’s test bench for inference, document processing, model adaptation and agent boundaries. When we share results, we say what was tested and how. A result on our bench is a starting point; your task still gets its own test.