Ask · Plainy

How can we help you today?

Plainy knows the studio: the people, the projects, who worked on what, and how Little Plains runs an engagement. Ask it anything you'd ask the team.

The persona lives in the weights. No system prompt, no retrieval.
First message wakes a GPU. Cold start ≈ 4 min; after that it's fast.

Instrument panel--:--:--

Model

checkpoint

…

base

…

adapter

LoRA r16 · SFT

renderer

…

system

none (in weights)

sampling

…

Endpoint

host

…

state

cold · not contacted

elapsed

—

remaining

—

imports — runtime
volume_read — registry
engine_init — weights + compile
adapter_warm — LoRA

Last turn

ttft

—

reasoning

—

answer

—

throughput

—

wall

—

finish

—

Session

turns

context

0 / 8192 tok

02k4k6k8k

Principles

Expose state and inner workings.
Explicit is better than implicit.
Verbosity over opacity.
Don't infantilize users.

Prototype: a post-trained persona model for Little Plains, built by Generative Experiences. Temporary. Counts marked ≈ are streamed deltas, not tokenizer counts.