∴ Praxilogue

A studio practice for both human and machine thinking.

I don't reach for AI for answers. I build with it to find out what changes when reasoning becomes shared rather than solitary, and where the line between human judgment and machine autonomy should actually sit.

Praxilogue is where that inquiry gets built, not just theorized. Each project tests a different edge: where dialogue sharpens a decision, where autonomy can run unattended, where a prototype earns the right to become something real.

01 Methodologies

Three ideas underpin everything here.

Progressive Context Building

Understanding compounds. Context accumulates through sustained interaction rather than being specified upfront, and the pattern reveals itself as it builds.

Co-Channeling Architecture

Interaction designed for both sides at once. Not just better prompting, but a structure where human and machine each show up sharper.

Socratic Dialogue

Questions before instructions. Reasoning surfaces through structured inquiry rather than being handed down.

02 Projects

Where the inquiry gets tested.

  • 01
    Cognitive Council↗

    Socratic dialogue, AI generated. A way to stay in the question longer.

  • 02
    Lexi↗

    An autonomous lexicographer. Runs its own daily sweeps, unsupervised.

  • 03
    Rêvesvoir↗

    A sequence of reflective exercises, AI synthesized. Ends by reading back the question you were actually asking.

  • 04
    Valence↗

    An interactive poem, worn like a watch. Proof the practice isn't only utilitarian.

  • 05

    Claude @ Home in design

    A smart home run on judgment, not just triggers. Home Assistant and Frigate, with Claude reasoning over what the sensors see.

03 On Augmentation

Building the scaffolding beforehand.

Being AI-augmented isn't reaching for a ready-made assistant when the moment calls for it. It's building the scaffolding beforehand: prompts, workflows, checkpoints that make judgment sharper before the moment arrives.

That instinct doesn't stay personal for long. The same question, what should stay human and what can run on its own, is the one every team hits once AI moves from novelty to infrastructure. Scaling it means the same thing at any altitude: give people something to build with, not just something to ask.

04 On Governance and Judgment

Knowing where autonomy should stop.

Every project here answers a question enterprises are quietly asking too: when does a prototype earn trust, and who decides? The interesting failures aren't technical. They're structural. A demo dies not from bad output but from no pathway to production. A sandbox becomes shadow IT not from bad intent but from no visible cost. A team disengages not from the tool but from training pitched at the wrong altitude.

The guardrails matter as much as the capability. Knowing where autonomy should stop isn't a constraint on the work, it's part of the design. Ethics isn't a review gate bolted on at the end; it's a variable in the architecture from the first sketch. Building unsupervised agents means designing the checkpoints where human judgment actually matters, the same discipline that separates a Lab that ships reusable assets from one that just accumulates demos.