Swarochish Chekuri

Position

I work on reducing the distance between intent and execution.

Three to four weeks became 53 seconds.

A question typed in plain English came back with the analysis, the models and the insights. Across the datasets we deployed it against, the answer arrived in under 53 seconds.

The distance still did not close. Somebody still had to know which question to ask, and what to do once the answer arrived. Compressing the work is not the same as removing the gap between wanting something and having it done.

What I am building now

Daemon AI: What if saying what you want was enough to make it happen? Products that reduce the work between an intention and an outcome. Personal assistance, personal-brand management and specialised professional systems are already in development. Building privately · incorporated 2026
Mnemos AI: What has to exist around intelligent models before we can trust them with more of the work? A deeper systems project I have been thinking about for several years and started building formally in 2026. Most of the work remains private. Building privately · incorporated 2026

What I have been working through lately

memory that survives the conversation
procedural knowledge for machines
governed autonomous execution
model-independent agent systems
AI that can build software
interfaces for systems that increasingly act rather than wait

Earlier I built Penser Analytics and Vipani. The work

If something here overlaps with a problem you are working on, write to me. I am particularly interested in people thinking seriously about autonomy, trust, and what has to be true before a system is allowed to act on your behalf. swarochish.c@gmail.com

Open questions

How much autonomy can a system earn, and what does it have to show first?

The more consequential the action, the more permissions, observability, determinism and reversibility it needs.
If the interface of the future is not a chat box, what is it?

Chat is good for ambiguity and bad at nearly everything else. The interface should emerge from the state of the work.
What is an agent actually bounded by?

Not the model. What it can sense, remember, understand and verify. Agentic systems are data systems.
What is an application, once it stops being the thing you open?

Applications look increasingly like an interface abstraction rather than the end state.

All questions, with positions and one that closed

Essay

The distance between intent and execution

Why the gap is hard, what an analytics product taught me about compressing it, why 53 seconds was not enough, and what an application becomes responsible for when the person stops operating it step by step.

Research

Working papers. Provisional by definition, versioned, and open to correction.

Working paper · v0.2

Trust Across Boundaries

An eight-layer model of trust with strict dependency ordering, spanning human, agent-human and agent-agent interaction, and where the unsolved problem actually sits.

Working paper · v0.2

Broken at the Seams

Multi-agent systems fail at the points where control, context or commitment crosses a boundary. Ten design principles for that layer, and a metric that would show the argument to be wrong.

Working paper · v0.1

A General Theory of Human-System Interaction

Felt ease treated as a regulation problem rather than a design problem. Complexity is allocated rather than removed, and the audit question is who absorbed what was taken away.

A few notes

Note 02