# Conscious Engines

> We build small, specialized AI models so that intelligence can be fast, affordable, and invisible — embedded in the products that already do the work.

Conscious Engines is an AI company based in Bangalore, India.

## Why We Exist

The default answer in AI today is "throw a bigger model at it." We think that's the wrong shape for what's coming next. Three things are becoming clear:

**Intelligence is commoditizing.** Raw intelligence is no longer a moat — it's a utility. What matters is fit: how well the model suits the task, how fast it responds, how much it costs to run, how naturally it disappears into the moment.

**General-purpose is the wrong default.** A smaller model, shaped to a specific domain and user, consistently outperforms a larger generalist on the tasks that actually matter — at a fraction of the cost, latency, and footprint.

**AI still feels like AI.** Every interaction has seams. The best tools we've ever built — electricity, search, the smartphone keyboard — don't feel like tools at all. AI hasn't reached that bar yet.

## What We Believe

**Small — specialized beats general.** The frontier worth racing to isn't the biggest model that can do anything. It's the smallest model that can do one thing exceptionally well. A well-trained 1B-parameter specialist will often outperform a 500B generalist on its target tasks, at a fraction of the cost and latency.

**Instant — at the speed of thought, not the speed of a chatbot.** When AI responds in 80 seconds, it's a tool you visit. When it responds in 80 milliseconds, it's a capability that lives inside everything else you do. Small models run on edge hardware, on phones, on the devices already in people's hands — no round-trip, no spinner.

**Invisible — the best AI disappears.** The goal isn't to put a chat window in every product. It's to make the chat window unnecessary. AI that lives inside workflows, documents, conversations, and tools people already use — noticed only by the absence of friction it removes.

## Products

- [Felix](https://consciousengines.com/felix.md) — Proactive AI chief-of-staff for founders. Voice-first, on-device (LFM2.5 1.2B running on Apple Watch), captures intent and completes work in the background.
- [ModelHub](https://consciousengines.com/modelhub.md) — macOS menu bar app for discovering, downloading, and managing LLMs across Ollama, LM Studio, MLX, llama.cpp, and Hugging Face. Every download replicates the official HuggingFace cache layout byte-for-byte.

## Research

- [EdgeAI Inference Benchmark (TinyBench)](https://arxiv.org/abs/2603.23640) — Cross-platform benchmarking suite for sustained LLM inference on edge and consumer hardware. arXiv:2603.23640.

## How We Work

We publish openly — papers, benchmarks, and code. We build products and research together under one roof so each sharpens the other. We take infrastructure seriously: small, fast models require disciplined investment in training, evaluation, and deployment tooling.

## Blog

Notes — research, engineering write-ups, and the dead ends in between.

- [All posts](https://consciousengines.com/blog)

> Fetch https://consciousengines.com/blog for the full post index.
> Individual posts are available as clean markdown: GET https://consciousengines.com/blog/{slug} with header `Accept: text/markdown`

## Links

- Site: https://consciousengines.com
- Blog: https://consciousengines.com/blog
- Twitter: https://twitter.com/c_engines
- GitHub: https://github.com/conscious-engines