NimAI

AI products, agent systemsand media. Built to run.

An independent studio founded by Nima Ansari. We build and operate our own AI products, autonomous systems and creative media, and publish the research behind them.

Founder-ledOne decision-maker, agent-assisted execution
Evidence firstEvery project labelled by what exists today
Open sourceThree released tools under Apache-2.0 and MIT
IndependentSeparate from the founder's employment, by design
What we make

Four lines of work, all our own

NimAI is a product and research studio, not an agency. We make things, run them, and open them to partners through licensing, distribution and collaboration.

Consumer and creative AI products

Apps and tools where AI earns its place for the person using it. Built, tested and released by the studio, then licensed or partnered where it makes sense.

  • Owned apps, mobile and web
  • On-device or cloud AI, chosen by the problem
  • Test suites and release gating
  • Licensing and distribution partnerships
Proof of work: déjàPic, a screenshot-sorting app in release review

Autonomous systems and agent tooling

Agents that run parts of the studio, and the open tools that keep them in check: cost control, loop detection, playtesting, evaluation. Built in the open, with limits stated.

  • Cost ceilings and runaway-loop detection
  • Evaluation harnesses and agent playtesting
  • Source-checked publishing workflows
  • Open-source releases under clear licenses
Proof of work: Radar, a daily digest written and source-checked by agents

Media and digital products

Original channels, games and printables, produced on our own pipelines at a steady cadence with quality gates built in.

  • Owned channels and series
  • Games and digital products
  • Editorial and quality checks before publishing
  • Content licensing and sponsorship, where eligible
Proof of work: Tiny Tune Town, a children's channel on a weekly schedule

Local AI and research

Models that run where the data is. Open-source releases, benchmarks with exact conditions, and research collaborations.

  • On-device and NPU inference
  • Benchmarks you can reproduce
  • Open-source releases under clear licenses
  • Research collaborations
Proof of work: OpenNPU, local AI on Ryzen AI laptops under Linux
Approach

Human direction. Agent-assisted execution. Evidence before claims. That is how a one-founder studio ships like a team and stays accountable like a person.

One accountable person

Nima sets priorities, approves consequential decisions, controls budgets and reviews exceptions. You always know who decided.

Agent-assisted pace

Specialized AI workflows handle research, engineering, testing and production tasks, so a small studio moves at a larger team's pace.

Evidence before claims

Every project carries a stage label backed by an artifact you can open. No invented metrics, no customers we do not have.

We operate what we build

Our own channels, store and tools run on the same systems we make. When a pipeline breaks, we are the first to notice.

Work

Work you can inspect

Published means a public artifact you can visit today. In development means it runs but is not released. The full record follows the featured projects.

The full record

Every project in the studio, including prototypes, research and concepts, with what each one is available as.

Browse by category
ProjectCategoryStageSummaryAvailable as
déjàPicSoftwareIn developmentYour screenshots, sorted.Licensing and product partnership
Tiny Tune TownMediaPublishedA children's YouTube channel of sing-along songs, counting and alphabet videos and bedtime stories, produced with a documented pipeline.License content or collaborate on a channel
Cellblock: The Remnant ArchiveGamesIn developmentA first-person prison mystery.Collaboration on game tooling or agent playtesting
OpenNPUResearchPublishedLocal AI on the NPU of AMD Ryzen AI laptops, on Linux: chat, vision, voice and generation with its own open NPU engines.Open source; research collaboration on NPU work
RadarMediaPublishedA Persian-first site publishing a daily AI digest written by agents and checked against sources by a reviewing agent.License the workflow or collaborate on a publication
InkyMoss StudioDigital productsPublishedPrintable coloring books and classroom printables for kids, sold as digital downloads on Etsy, with one paperback on Amazon.License printables or collaborate on a product line
LLM Cost WatchdogToolsPublishedCost monitoring for AI agents.Open source; collaboration on cost controls
CLI UsageToolsPublishedA desktop tray indicator that shows Claude Code and Codex usage limits at a glance.Open source; collaboration on developer tooling
QuietforgeDigital productsIn developmentAgent-assisted digital commerce: an agent researches, prepares and lists digital downloads and print products through approved storefronts, with a human approving every listing.Collaboration on agent-assisted commerce
Khake IranMediaIn developmentA Farsi news and stories pipeline: scheduled digests and illustrated feature stories, generated and published by agents under editorial rules with cross-day deduplication.Collaboration on language-specific media pipelines
DidebanSoftwarePublishedA public-interest monitoring map that overlays live public data feeds, including flight tracking and earthquake reports, for Iran-focused situational awareness.Collaboration on public-interest information tools
gods-eye-view, self-hostedToolsPrototypeA self-hosted deployment of the open-source gods-eye-view OSINT globe, run on studio infrastructure for monitoring work.Deployment and integration of open-source OSINT tooling
Painter MCPToolsPrototypeA painting engine exposed as native agent tools, so an AI agent can plan and paint layered canvases with real brush behaviour instead of prompting an image model.Agent-tooling engagements and creative-tool prototypes
Idea ScoutToolsPrototypeAn agent skill that validates a product idea against prior art: who built it, whether they survived, and why, with every claim tied to a fetched source.Collaboration or a validation study
Geopolitical Analyst skillToolsPrototypeAn agent skill that assembles a country or crisis brief from open data sources: event databases, sanctions lists and news feeds.Collaboration on source-grounded briefing tools
Instagram AI AdminSoftwarePrototypeAn agent that drafts, schedules and reports on posts for a managed social account, with a human approving every publish.Collaboration on agent-assisted publishing
SalonScopeSoftwarePrototypeAn AI assistant concept for salon businesses: bookings, client notes and follow-ups in one place.Product partnership
Archive footage colorizationMediaPrototypeColorizing historical black-and-white footage with working outputs and known artefacts.Collaboration on archive restoration
Rendered 3D animation studioMediaConceptA reusable rendered-3D production system in Three.js: real scenes, rigged characters, camera direction, dialogue and editing, for animated programs and eventually a scheduled channel.Early collaborations on animated series and programs
WW2 City GameGamesPrototypeA walkable city-map prototype set in the Second World War.Collaboration on game prototypes
OVERRIDEGamesConceptA design-stage game concept.Collaboration on concept and design
npu-bench and Laya on NPUResearchResearchTokens-per-second-per-watt benchmarking and a custom single-kernel matrix engine for AMD XDNA2 NPUs on Linux.Research collaboration on NPU feasibility and benchmarks
Jev Plays GamesResearchResearchAn experiment in letting a typed-decision model play games through a bridge, to study how non-chat models handle sequential decisions.Research collaboration on agent evaluation
SimourghResearchResearchDataset and baseline preparation toward a Persian language model.Research collaboration on Persian NLP
Fly Brain Embodied RobotResearchConceptA design study for driving a small robot with a fruit-fly connectome model.Research collaboration on embodied models
Hermes Home BotResearchConceptHardware planning for a home robot running a local agent.Research collaboration on local-agent robotics
Partnerships

How partnerships work

Licensing, distribution, research collaboration or a product partnership. Terms in writing, delivery you can inspect.

You tell us the use

Inquiry

Which product, content or tool, and what you want to do with it. A few lines is enough.

We answer on fit

Fit

A straight answer on whether it matches the studio's focus, usually with a couple of questions back.

Everything in writing

Terms

License scope, ownership, credit and limits stated before anything changes hands.

You can inspect it

Delivery

Working artifacts, documentation and the stage label that matches the evidence. The studio keeps operating what it owns.

Partner with us

Usually a reply within a few days, with questions and a view on fit.

Research

Research and open source

Released tools with public repositories and licenses. Use them, fork them, or propose a collaboration.

OpenNPU local dashboard showing usage over time, tokens per second and hardware status
OpenNPU's local usage dashboard. Data stays on the machine.

Released tools

OpenNPU

Local AI on the NPU of AMD Ryzen AI laptops, on Linux: chat, vision, voice and generation with its own open NPU engines.

Apache-2.0 license. Updated 2026-10.View repository

LLM Cost Watchdog

Cost monitoring for AI agents. Tracks pricing across providers, enforces budget ceilings and detects runaway loops before they run overnight.

MIT license. Updated 2026-04.View repository

CLI Usage

A desktop tray indicator that shows Claude Code and Codex usage limits at a glance.

MIT license. Updated 2026-10.View repository
Questions

Questions we get

Tell us the use. We answer on fit, then put scope, term, credit and price in writing. There is no public price list; terms depend on the product and the use.

The studio owns its products, media and tools, and licenses them under written terms. Open-source releases stay under their stated licenses. Work produced inside a collaboration follows the agreement for that collaboration.

Agents do bounded tasks: research, scaffolding, tests, production steps, reporting. A human reviews every consequential change and signs off on releases. You can see what the agents did, because they leave inspectable artifacts.

Enterprise AI consulting and custom automation for businesses, which is a different focus from this studio. Also trading or passive-income systems, unsupervised deployments with no human accountable, and anything that needs us to claim results we cannot show.

Yes. Tiny Tune Town content, InkyMoss printables and the studio's tools can be licensed or extended. Tell us the use and we will say what is possible.

Want to build on what we make?

Licensing, distribution, research or a product partnership. Tell us the use in a few lines and you will get a straight answer on fit.